# Xivic Insights (full text) > Full text of every Xivic insight article, concatenated in one file for language-model retrieval. Each article section is prefixed with its canonical URL. For a lighter, linked index see https://www.xivic.com/llms.txt. --- # The Value Friction Index Source: https://www.xivic.com/xivic_insight_value_friction_index.html > The Value Friction Index | Xivic Insights | insights and frameworks from Xivic. ## The Problem: Eight Pilots, Zero Clarity The CTO emails you the monthly AI dashboard. Eighteen projects. Six months of runway. Three separate tools logging results. You pull the file into your calendar. Your CFO asks what this means for EBITDA. Your operating partner asks which pilots to fund. You have no answer, not because the data is missing, but because you're measuring inputs (pilot count, budget spent) instead of outputs (where value actually lives). This is the pattern across most PE portfolios. Companies run AI initiatives as isolated experiments. There's no common language for what's working. There's no diagnostic to separate the pilots that move the needle from the ones that feel good. ## Why Your Existing Measures Fall Short Most portfolio companies rely on ROI-per-project to justify AI spend. The problem is obvious: not every AI initiative produces revenue, and not every dollar saved goes to the bottom line. A 40% reduction in data-storage costs looks good until you ask whether it compounds over time or vanishes with the next infrastructure decision. Some portfolios try revenue-share-of-AI, the contribution AI initiatives make to top-line growth. But this treats each project as independent. It misses the fact that real value in a mature portfolio comes from **compounding**: when AI improves operations, which fuels revenue, which attracts better talent, which drives better decisions. Measure the parts and you miss the whole. Vanity metrics, pilot count, headcount in AI, investment dollars, are even worse. They tell the board that you're doing something without telling you whether it matters. ## Introducing the Value Friction Index The Value Friction Index is a diagnostic scorecard that measures the **friction between strategy and execution** in how a portfolio company deploys AI and data assets. It scores five value levers, each on a scale of 1 to 5, based on observable evidence in your operating model, product, and financials. It's not an audit. It's a compass that points you toward where stuck value lives. The five levers are: - **Revenue Expansion**: are data and AI initiatives directly driving customer acquisition, conversion, or wallet share? - **Margin Improvement**: is automation or optimization reducing cost-to-serve without sacrificing quality or customer experience? - **Speed & Throughput**: are workflows, sales cycles, or time-to-market measurably faster because of AI or data integration? - **Risk Reduction**: is the company better able to predict, avoid, or mitigate operational, financial, or reputational risks? - **Multiple Expansion**: are the business model shifts unlocked by these capabilities visible in valuation multiples or customer stickiness? ## How VFI Is Scored Each lever gets a score from 1 to 5. You don't rely on self-reporting. You look at the business. A score of 1 means the lever is fragmented or aspirational. There's a pilot, but no evidence that it's flowing into the core business. A 5 means it's woven into the operating model. The mechanism is repeatable. The results compound. For Revenue Expansion, that means looking at whether new revenue attributable to AI-driven features is growing quarter over quarter and flowing into core product metrics. It means checking whether your marketing team uses data to segment, personalize, and convert, or whether they run campaigns the same way they always have. For Margin Improvement, you're looking at cost-per-transaction, cost-per-employee-output, or cost-of-goods-sold in categories where automation was deployed. Is the improvement persistent? Does it come from a process change or just a one-time reduction? Speed & Throughput is the most concrete. You're measuring cycle time: how long from inquiry to estimate? Quote to close? Idea to launch? You should have these numbers in your board deck. If you don't, you don't have clarity on execution friction. Risk Reduction is harder to quantify, but it shows in churn, ARR volatility, or customer-concentration metrics. It shows in claims frequency or underwriting error rates. The diagnostic is: did we move the needle on something that matters to the business, or just collect more data? Multiple Expansion is the ultimate signal. Are comparable companies with similar AI maturity trading at higher revenue multiples, EBITDA multiples, or retain-rate multiples? Are your customers stickier because of what you've built? None of these are self-reported. You ground each score in the books, the systems, and the evidence your operating partners can actually see. ## What High-VFI Looks Like A high-VFI portfolio company doesn't have more AI pilots than anyone else. It has fewer, and they're deeper. The operating model is unified around data. There's a single source of truth for customer behavior, product usage, or operational metrics. When the product team wants to run an experiment, they don't build a new dashboard, they query the platform. When the supply chain team wants to optimize, the data asset is already there. When revenue decides to target a new segment, they know the margin profile and the churn risk before they spend a dollar. The playbooks are durable. When a new person joins the revenue team or the ops team, they inherit a repeatable way of using data to make decisions. It's not dependent on one person's SQL skills. It's not dependent on the current data analyst staying. The capability persists. The data assets compound. Each successive AI initiative doesn't start from zero. It leverages what came before. The data model strengthens with use. The integrations expand the surface area for the next innovation. Over 18 months, you move from eight isolated pilots to one coherent system that powers continuous improvement across the whole business. That's high friction *elimination*. And it shows in the financials. ## How to Run Your First VFI Scan Start by picking one portfolio company. It doesn't need to be your most mature AI player. It needs to be one where you're spending operating time right now. Schedule a half-day with the CEO, CFO, and the operating partner who knows the business best. Don't do this over email. You need to ask follow-up questions, and you need to see the organization's actual answer, not a PowerPoint answer. Walk through each lever. For Revenue Expansion, ask: "What revenue has AI or data driven in the last 12 months?" You should get a number. If you get a story, you don't have a clear mechanism yet. For Margin Improvement, ask to see cost-per-unit or cost-per-transaction trends. For Speed & Throughput, ask for a specific cycle-time metric. You probably have it already, it's just not labeled as an AI outcome. Then, for each lever, score it 1 to 5. Don't average. Note where the friction is highest. That's usually where the biggest value creation opportunity sits. A portfolio company with a 1 or 2 on Revenue Expansion probably has a clear path to unlocking value by reorganizing how the product or marketing org uses customer data. A company with a 3 on Speed & Throughput usually has the infrastructure but not the discipline to apply it everywhere. Create a simple one-pager: company name, the five scores, and three near-term moves to close each friction point. Share it with your operating partner. This becomes your diagnostic and your roadmap. ## The Compounding Effect VFI isn't a goal. It's a measurement. But measurement drives behavior. When you score your portfolio companies against these five levers, you start to see which ones are building durable, compounding value and which ones are running experiments. You start to allocate operating time to the right friction points. And in exit conversations, you can point to something tangible: "This company moved from a 2.8 VFI to a 4.2 VFI in 18 months. That's why the multiple went from 8x to 11x EBITDA." Value that's compound is value that sticks. It's value that survives a leadership change, a market cycle, or a new competitor. That's the kind of value that shows up in exit multiples. --- # The PE Value Creation Playbook Has Changed Source: https://www.xivic.com/xivic_insight_pe_value_creation_playbook.html > The PE Value Creation Playbook Has Changed | Xivic Insights | insights and frameworks from Xivic. ## The Old Playbook Broke. Here's Why. For twenty years, private equity's value creation math was consistent: buy a business at 6x EBITDA, install an operating partner, cut costs by 15-20%, grow EBITDA by 3-5% annually through modest top-line expansion, and exit at 8-9x. The leverage and multiple arbitrage alone delivered a 2.5-3.2x money multiple. It worked because rates were low, multiples were fat, and you could hold for four to five years. That era is over. Structurally higher rates-at 5-6% now versus 2-3% in 2019-have compressed entry and exit multiples by 25-40%. Median PE hold periods have stretched from 4.5 to 6.7 years. More critically, financial engineering now accounts for a shrinking slice of total returns. Operating improvements, organic EBITDA growth and margin expansion, now drive 47% of buyout value creation, up from 28% a decade ago (Bain, 2024). The firms that raised capital in 2021-2022 are learning this lesson painfully: you can't buy your way to target IRRs anymore. LPs have noticed. They're asking a question that would have seemed absurd in 2018: "What is your operating model?" They want to see repeatable, scalable processes that can be applied across a portfolio to create durable economic lift, not one-off transformation initiatives that die when the consultants leave. The firms winning now understand something fundamental: the unit of value creation is no longer a single portfolio company. It's the entire portfolio as a platform. ## Three Eras of PE Value Creation The evolution is clear in retrospect. **Era One: Financial Engineering (1990s-2008).** Leverage was the lever. You bought a stable business with reliable cash flow, loaded it with debt, and let the equity multiply as you paid down principal and captured spread. EBITDA growth was nice but not essential. This worked until it didn't, 2008 made that clear. **Era Two: Operating Improvement (2010-2019).** Post-crisis, firms learned to actually run their businesses. New management teams, cost discipline, modest organic growth, operational metrics dashboards. The best firms, Carlyle's operating group, Roark Capital's vertical integrators, built repeatable playbooks. A single operating partner could unlock 3-5 percentage points of margin or 2-3% organic EBITDA growth if they knew what to do. This was real value. But it was still PortCo-by-PortCo. The learning from one business didn't systematically transfer to the next. **Era Three: Portfolio-Wide AI Operating Systems (2024-present).** The leaders now are building shared data substrates and reusable operating infrastructure that every PortCo plugs into. They're not hiring consultants to analyze each business separately. They're installing operating systems, data layers, agentic workflows, measurement frameworks, that apply across the portfolio and compound in value over the hold period. The infrastructure itself becomes transferable IP at exit. This shift is not theoretical. Firms like Roark Capital, which owns Driven Brands, a Xivic client, are already running this playbook. They don't install a digital transformation at Vroom and then repeat it separately at Heydude. They build once, apply everywhere, measure systematically, and compound. ## Why Single-PortCo Transformation Is the Wrong Unit of Analysis Most PE firms still approach digital and AI capability-building as a PortCo problem. "We're going to hire a transformation partner to upgrade the IT stack at this business." That's backward. It's expensive, it's slow, and it doesn't transfer. Consider the mechanics. A single PortCo transformation takes 6-12 months, costs $500K-$2M depending on scope, and creates tacit knowledge locked in one organization. By the time the next PortCo acquisition lands, the learning is stale. You restart. Six months later, you realize the third PortCo has similar problems but a different tech stack, so you call a different firm. By year three of your hold period, you've spent $4-6M across the portfolio and still don't have a coherent operating system. Now consider the alternative: a portfolio-wide AI operating system installed once and shared across all PortCos. It has a shared data layer so that every business feeds consistent metrics, customer cohort health, unit economics by channel, margin drivers, throughput velocity. It has reusable agentic workflows that automate common processes: demand forecasting, inventory optimization, personalization at scale, cost anomaly detection. It has a measurement framework, Xivic calls it the Compound Value Model, that tracks durable lift across quarters and shows what's working across the portfolio. New PortCo acquisition? Onboard in 2-3 weeks instead of 2-3 months. The infrastructure is already there. The playbooks are proven. You don't reinvent; you apply. ## What a Portfolio-Wide AI Operating System Actually Looks Like Let's be concrete. A functioning system has five layers: **Shared Data Substrate.** Every PortCo pipes consistent data into a cloud-based warehouse. Revenue by customer, product, channel, and period. Costs by category and center. Customer acquisition, retention, and lifetime value metrics. Inventory and fulfillment metrics. Not every metric is relevant to every business, a SaaS PortCo doesn't care about inventory turns, but the framework is standardized. This takes 4-8 weeks to set up per PortCo once the infrastructure exists. **Diagnostic Engine.** Automated scans run monthly across the portfolio using the Value Friction Index, a diagnostic framework that scores where value is trapped across five levers: Revenue Expansion, Margin Improvement, Speed & Throughput, Risk Reduction, and Multiple Expansion. The VFI doesn't recommend actions; it surfaces friction. "This PortCo has excellent margin discipline but is leaving 12% on the table in throughput velocity." Another PortCo is strong on velocity but weak on customer expansion. The diagnostic is consistent across 30+ portfolio companies. **Reusable Agentic Workflows.** Once you've identified friction, you deploy agents, AI systems with clear decision authority and operational boundaries, to address it. An agent trained on your best PortCo's demand forecasting process can apply that logic across similar businesses. Another agent handles customer segmentation and personalization. Another detects cost anomalies and escalates them. These agents run 24/7. They're not consultants; they're infrastructure. **Velocity Operating System (VOS).** This is your delivery layer. It's not a tech platform; it's a method for shipping operating improvements in weeks instead of quarters. When the VFI flags "this PortCo is weak on customer retention," the VOS deploys a playbook, data queries, agent workflows, measurement dashboards, that addresses the friction, measures the lift, and documents what worked. The next PortCo with similar friction gets the proven playbook in days. **Measurement and Compounding.** None of this matters without a system that measures and locks in durable lift. The Compound Value Model tracks the impact of each operating improvement, isolates it from noise, and shows how it compounds over quarters. You can see not just that you improved margin, but how much was durable, how much transferred when you applied it elsewhere, and what lift you're still leaving on the table. ## The Multiple Expansion Math: Why Transferable Infrastructure Beats One-Off Transformation Here's where the financial lever clicks into place. A traditional PortCo transformation, new management, cost cuts, process improvements, might lift EBITDA by 8-12% over the hold period. That's valuable. It moves you from 6x entry to 6.8-6.9x exit valuation. A modest boost to the multiple. Now introduce a portfolio-wide AI operating system. The same improvements are available, but they're not limited to one PortCo. The VFI runs across your whole portfolio. It shows you that five of your ten PortCos are underperforming on customer cohort retention by 20-30% versus the best-in-class performer. The agentic workflows that fixed retention at PortCo A deploy to PortCos B, C, D, and E. The lift isn't isolated to one business; it's portfolio-wide. The compounding is faster. More crucially, here's what happens at exit: the infrastructure itself has value to the buyer. In a traditional exit, the buyer acquires the improved business but the consultant knowledge, the dashboard framework, the process improvements, those evaporate. The buyer hires their own operating team and starts over. In a portfolio-wide AI system exit, the buyer acquires a transferable operating system. They can apply it to their existing portfolio. They can extend it. They can immediately reduce the integration friction that usually eats 3-5% of synergy value because the measurement and process infrastructure is already there. That transferability has a price. It can add 0.5-1.0x to your exit multiple. It can expand your buyer universe to include buyers who value operating leverage, not just financial buyers looking for multiple arbitrage. ## How to Start: Run the VFI Scan You don't transform overnight. You start with clarity. Run the Value Friction Index diagnostic across your 3-5 most mature PortCos. Score them across the five levers. You'll see immediate patterns: which businesses are underperforming on which dimensions, which have solved problems that others haven't, where the highest-leverage shared opportunities exist. This work takes 6-8 weeks and costs $150K-$250K depending on data maturity. Most firms find 2-3 high-priority shared friction points worth addressing across the portfolio. Pick the highest-leverage one. Deploy a proof-of-concept agentic workflow to address it. Measure the lift at the first PortCo. Replicate to the second. Document. Lock in the lift. You'll have a repeatable playbook inside of three months. Scale from there. Quarterly, you expand the system: new data sources, new workflows, new PortCos. By year two of the hold period, you have systematic operating leverage across the entire portfolio. By year three, it's compounding. ## Your Portfolio Is a Platform, Not a Collection The PE firms that will win the rest of this decade are those that stop asking "how do we improve this business?" and start asking "how do we operate this portfolio as a single machine?" It's a different question. It demands different capabilities. It demands infrastructure you build once and apply systematically. Xivic was born into this era. We've spent the last two decades in digital operating, we know how to install systems that scale. We've worked with Roark Capital, Carlyle, and growth-stage brands across 80+ operating challenges. We've learned where the friction lives and how to measure durable lift. We don't write decks. We install infrastructure. We don't recommend transformations; we deploy them. We build the systems that let your portfolio compound value across the hold period and transfer that value to your exit buyer. The playbook has changed. The firms moving fastest understand: the moat isn't in finding better deals. It's in operating better. *Xivic is an AI-first value creation operating partner for private equity firms, enterprises, and growth-stage brands. We build portfolio-wide operating systems that compound durable value across the hold period.* --- # Digital Diligence Source: https://www.xivic.com/xivic_insight_digital_diligence.html > Digital Diligence | Xivic Insights | insights and frameworks from Xivic. Private equity firms have spent the last decade professionalizing digital diligence. They hired technical advisors, built playbooks, learned to score cloud adoption and legacy modernization risk. By 2026, most mature PE shops have a working protocol: confirm the target runs on ERP, has some cloud spend, maintains a website that doesn't crash, and call it done. That protocol was never wrong. It just became incomplete. The underwriting question has shifted. It's no longer "Does this business have a functioning tech infrastructure?" That's table stakes. The real question now-the one that determines 300-500 bps of value creation-is this: **Can this business absorb AI operating leverage over a 5-7 year hold?** Those are fundamentally different questions. A company with a 15-year-old legacy ERP but pristine transactional data, well-documented workflows, and a CEO who's already piloting AI tools is a radically better AI-era target than a company with a modern SaaS stack, siloed data, and a leadership team viscerally opposed to anything that feels like automation. Traditional digital diligence would score the latter higher. It would be catastrophically wrong. This article introduces a 90-day framework, the diagnostic and timeline that PE deal teams can run, in parallel with legal and commercial diligence, to measure AI readiness and identify the specific value creation wedges worth 6-8 figures on day one of ownership. ## Why Traditional Digital Diligence Misses the AI-Era Upside Most PE firms budget $80K-$150K for tech due diligence. A third of that goes to a third-party tech audit (security, cloud, architecture assessment). Another third goes to a few advisor calls. The rest is the operator's seat-of-the-pants judgment. That spend isn't wasted. But it's optimized for the wrong thing. It answers: "What's the technical debt?" It doesn't answer: "What's the technical *opportunity?*" Here's the gap: a business can have zero technical debt and still be worthless as an AI transformation target. Conversely, a business can be loaded with legacy code and highly valuable, if the underlying data and processes are clean. The mistake, in concrete terms: traditional diligence looks at tech stack modernity. It scores Kubernetes and microservices and headless commerce and modern data lakes. These are genuinely useful things. But they're not correlated, not even loosely, with AI readiness. You can have a Kubernetes cluster managing a pipeline of dirty, aggregated, duplicated data. You can have a 2005 system that produces a single, clean source of customer truth. AI readiness depends on four things. None of them are in the standard tech audit. **Data flow integrity:** Can we see what the business actually does? Not kinda. Actually. Not via a reporting layer someone built three years ago and hasn't updated. Can we see the source transactions? Are customer records deduplicated? Can we trace margin from deal to delivery to invoice? If there's a mystery in the P&L, can we find it in the data? Firms with terrible tech stacks often have this. They compensate for bad software with obsessive spreadsheet discipline. That spreadsheet discipline is *gold* for AI work. **Workflow definability:** Are the core processes codified? Can you write them down? Not as religious artifacts, as actual sequences. If the answer is "it depends" or "the experienced people just know," the business is fragile *and* unaugmentable. A business where 80% of order fulfillment is rule-based (we do this if this, that if that, exception if this) is infinitely more valuable for AI leverage than one where it's 40% rule-based and 40% intuition and 20% "the person who knew quit." **Organizational absorbency:** Will your operators adopt the tools, or will they reject them? This isn't theoretical. We've watched a $200M manufacturing business reject a $2M warehouse optimization system because the ops VP felt threatened. We've watched a fragmented services firm absorb AI-driven scheduling in four weeks because the branch managers saw immediate relief. Leadership buy-in is usually binary. Does the CEO believe AI is opportunity or threat? (Skeptical is fine. Hostile is not.) Does the executive team see time savings or job loss? Do frontline operators think you're trying to automate them or augment them? This is culture, which most PE financial models still treat as unmeasurable. It's not. It's measurable and it's critical. **Economic unit cleanliness:** Can you measure unit economics? Not EBITDA. Unit economics. Margin per customer, per order, per employee-hour. Can you see cost of acquisition? Can you see delivery cost? Can you isolate the economics of a customer cohort, a product line, a sales channel? If you can, you can use AI to systematically improve each vector and measure the lift. If you can't, you're flying blind. You'll implement tools. You'll hope they work. You'll argue about whether things got better. These four dimensions, data, process, people, measurement, are what AI readiness actually means. A business strong in all four is a 20-30% EBITDA lift candidate. A business weak in all four is a 5-8% candidate, if you're lucky. A business with three strong and one weak has a ceiling you can identify on day one of ownership. ## The 90-Day Framework The framework divides into three phases, aligned with deal timeline: 30 days of signal gathering pre-LOI, 30 days of confirmatory diligence, and 30 days of post-close substrate install. ### Phase 1: Signal Scan (Days 1-30, Pre-LOI) **Objective:** Quick signal on AI readiness. Go/no-go on the deal's AI value thesis, plus a rough quantification of upside. **Effort:** Eight hours of your team's time, one working day of the target's senior management. **Method:** Call the CFO and COO. Here are the five diagnostic questions. Allocate 10 minutes per question. - **Data question:** "Walk me through your single source of customer truth. Is it your ERP? A data warehouse? A CRM that talks to your ERP? Or do you have three systems that don't fully talk to each other?" Listen for complexity and gaps. Write down the names and version years. Ask: "If I wanted to know the true margin on customer X's orders in 2025, where would that data live?" - **Process question:** "Take your biggest revenue-generating process, sales, delivery, support, whatever. What percentage of it is rules-based (you do this because of policy), what percentage is judgment, and what percentage is pattern-matching that you'd be hard-pressed to explain?" A business that says "75% rules, 20% judgment, 5% pattern" is codifiable. A business that says "40% rules, 50% judgment, 10% pattern" will be hard to augment. - **People question:** "If I told you we were going to implement AI tools to augment your operations team over the next year, not automate people, augment, what would be your team's stance? Excited? Skeptical? Resistant?" Excited is rare. Skeptical is normal and fine. Resistant is a red flag. Write down the name of the person most likely to resist and why. - **Measurement question:** "Can you tell me the unit economics of your business right now? Not EBITDA. Gross margin per customer? Cost of acquisition by channel? Delivery cost per order? What metrics can you actually see today?" Write down the ones they can see and the ones they can't. The ones they can't are your first 90-day upgrade targets. - **Regulatory question:** "Are there any regulatory or contractual constraints on how we can use customer or operational data? GDPR, HIPAA, customer agreements that restrict analytics or automation?" Write down every constraint. Some businesses can't do certain kinds of AI work. That's not a dealbreaker, it just changes the scope. **Output:** An AI Readiness Score on four dimensions (Data, Process, People, Measurement), a register of regulatory constraints, and a rough thesis on where the upside lives. Scoring: 1-3 on each dimension, where 1 is "significant constraint," 2 is "acceptable, with work," and 3 is "genuinely clean." A business that scores 3-3-3-3 is a 20-30% EBITDA lift candidate. A business that scores 1-2-2-2 is probably 6-10%. This takes a PE team maybe four hours of analysis and write-up. It's worth the clarity. ### Phase 2: Confirmatory Scan (Days 31-60, During Diligence) **Objective:** Lock in the value creation thesis with specific leverage points. Quantify AI upside by value lever. **Effort:** One week on-site, plus two weeks of remote data analysis. **Method:** This is where Xivic's Value Friction Index (VFI) comes in, but the logic applies regardless of who runs it. You're scoring the five value levers that drive PE returns: margin, growth, capital, risk, and multiple. For each lever, you assess the current state, the friction points, and the AI-specific opportunities. **Margin lever:** Where is the business losing margin today? Manual workflows eating hours? Repricing decisions based on incomplete data? Delivery routes run the old way? Use case: an industrial distributor where 30% of orders are repriced after quote because the system missed margin opportunities. AI-driven pricing, fed by your clean data and clean processes, closes that leak. Quantifiable upside on day 90. **Growth lever:** What's preventing faster growth? Sales team spending time on non-selling work? Manual RFP response? Lack of visibility into what actually closes? Use case: a B2B services firm where the sales team spends 40% of time on non-selling. Deploy AI-driven workflow augmentation, research, proposal assembly, follow-up sequencing, and redirect 20% back to selling. That's 1-2 points of margin, plus velocity. **Capital lever:** What capital is tied up in friction? Accounts receivable stretched because invoicing is manual? Inventory because planning is rule-of-thumb? Use case: a supply chain business that carries 60 days of inventory as a safety buffer. Predictive demand, enabled by AI, cuts that to 45 days. At $50M revenue with 40% COGS, that's $4M of freed capital. **Risk lever:** What risk is the business exposed to that AI can mitigate? Customer concentration because your data doesn't show it? Operational concentration in key people? Margin compression because you can't see competitive risk early? AI here is protective, not aggressive. But it enables faster adjustment. **Multiple lever:** What would make this business worth more in a secondary sale? Recurring revenue visibility? Lower customer concentration? Lower key-person risk? Better unit economics transparency? AI work on the first three translates to multiple expansion. The week on-site is spent walking the business, interviewing operators, pulling data, and stress-testing the signal-phase thesis. The output is a VFI score, a 1-10 across each dimension, and a ranked list of the 2-3 highest-leverage AI value plays. **Output:** Confirmatory VFI score, 2-3 specific use cases with quantified upside (in dollars and timeline), a 100-day operating plan that addresses the highest-leverage wedges first, and a Compound Value Model (CVM) that shows the expected AI-driven lift across the hold period. ### Phase 3: Substrate Install (Days 61-90, Post-Close) **Objective:** Build the technical and organizational foundation for sustained AI-driven value creation. **Effort:** 15-20% of the COO/CTO's time for 90 days. **Method:** This phase is not "full AI transformation." That's a 18-24 month program. This phase is "substrate install", the infrastructure that makes transformation possible. The substrate has three components: **Unified data layer:** Not a data lake. Not even necessarily a warehouse. The simplest version: a set of automated data pipelines that pull source transactions from your core systems, ERP, CRM, point of sale, payroll, into a single analytical space. Deduplicate customer records. Standardize codes. Create a clean business rules layer so that "revenue" means the same thing in every query. This isn't pretty. It's not a BI tool. It's a production dependency, and it's foundational. If you build AI tools on top of dirty data, they'll propagate garbage faster. Timelines: 60-90 days for a 10-15 person, $50M revenue business. Longer for bigger, messier companies. **First agentic workflow:** Pick the highest-leverage use case from the confirmatory phase and build it. Not a pilot. A production tool used by at least one team, every day, solving a real problem. It's usually something like: AI-assisted RFP response, AI-driven pricing recommendation, or AI-based scheduling. The point isn't the tool. It's the muscle memory. Your team learns what it feels like to work with AI-augmented workflows. They build confidence. They see lift. They ask for the next one. Timelines: 45-60 days from design to daily use. **VFI baseline lock:** By day 90, run your first post-close VFI. Lock in the baseline. Document the current state of data integrity, workflow definition, organizational absorbency, and economic unit clarity. This becomes your measurement anchor. Everything you do over the next 18 months is scored against this baseline, using the CVM dashboard. **Output:** Operational data layer live. First AI workflow in use. VFI baseline and CVM dashboard ready. You can now see, in real-time, where AI is driving lift and where friction persists. ## Red Flags That Should Repricing or Kill the Thesis Some diagnostics are deal-breakers, or at least repricing events: - **No single source of customer truth.** If customer data is fragmented across systems with no master record, you're starting from a hole. Fixable, but it's 6+ months of work before AI gets real traction. - **Leadership actively hostile to AI.** Skeptical is fine. Hostile is not. If the CEO sees AI as a threat to the business or their authority, antibodies will kill any initiative. - **Regulatory exposure on data handling.** Some businesses can't use customer data for analytics or AI work due to contracts or regulation. That's not a dealbreaker, it just changes the upside. - **Concentration risk in irreplaceable tribal knowledge.** If 40% of EBITDA is in the head of one person who refuses to document process, your upside is capped. ## Green Flags Worth Paying Up For - **Clean data.** If the business has already invested in data quality, you're ahead of the curve. That's worth points in your offer. - **Definable core workflows.** If the business has documented, rule-based processes, you can move faster. - **Leadership already experimenting with AI.** A CEO piloting ChatGPT with the team is often a signal that the culture is ready for deeper work. - **Customer segment where AI creates visible lift quickly.** Some businesses see ROI from AI in weeks (scheduling, customer service). Some take months. If you can see a fast-feedback loop, that changes your operating timeline. ## The Operating Thesis Diligence is no longer about what the business is. It's about what it can become. The 90-day framework is a discipline for seeing that future clearly, for quantifying it, and for building the operating substrate that makes it real. It turns AI readiness from a qualitative hunch into a measurable, staged program that you can staff, fund, and track. It also changes how you price the deal. A business with a 3-3-3-3 AI readiness score is not the same as a business with a 1-2-2-1 score, even if their LTM EBITDA is identical. The upside is different. The execution risk is different. The multiple you should pay is different. That's the point. This framework makes the difference concrete. And in PE, concrete is where value lives. --- # The Velocity Operating System: 5 Days to Deploy, Not 5 Quarters Source: https://www.xivic.com/xivic_insight_velocity_operating_system.html > The Velocity Operating System: 5 Days to Deploy, Not 5 Quarters | Xivic Insights | insights and frameworks from Xivic. *Enterprise AI doesn't fail in design. It fails in deployment velocity. A reusable engineering substrate is the difference.* ## The Longest Quarter A CTO sits across from her board in month eighteen of a three-year roadmap. The room is quiet except for the slide advance. On the screen: "AI Initiative Phase 3 - Model Evaluation Complete, Ready for Pilot Design." She's been saying some version of this for four quarters. Across the industry, a competitor in her space shipped their first AI feature in month three. Then another in month five. A third in month nine. Four production systems, each one faster than the last, each one generating measurable returns. The CTO's company has one pilot-still in evaluation, still uncertain whether it will ever reach customers. The difference was not talent, not budget, not strategy. The difference was substrate. ## Why Enterprise AI Stalls Enterprise AI projects do not fail because the models are wrong or the strategy is misguided. They fail because every single project starts from zero. The first failure mode is reinvention. An organization builds a RAG pipeline for use case A. Six months in, the platform team kicks off use case B, which needs a different retrieval layer because the data architecture differs. Use case C requires its own vector store tuning, its own embedding model, its own document chunking logic. By month twelve, the organization has funded three separate implementations of fundamentally the same problem. The engineering tax is invisible but devastating: every new project starts as a six-month tax before it can even begin adding value. The second failure mode is the evaluation gap. When each project has its own evaluation harness-or no harness at all, model changes become coin flips. Did we improve performance or just overfit to local data? Was that expensive fine-tuning worth the compute cost? Can we safely swap this model for a cheaper inference provider? Without a shared, calibrated evaluation framework that runs before and after every deployment, the team has no way to answer these questions with confidence. The project limps forward on intuition and expensive pilot phases. The third failure mode is production debt. Prototypes that worked in a notebook fail at scale. They lack observability, so you don't know when they're degrading. They lack guardrails, so unexpected inputs crash the system or hallucinate persuasively. They lack fallback logic for when models fail. The organization built a six-month prototype and spent the next three quarters hardening it. By then, the window of competitive advantage has closed. Each of these modes is solvable independently. Together, they are fatal. ## What the Velocity Operating System Is The Velocity Operating System is the reusable engineering substrate that collapses these three failure modes at once. It is not a product, not a platform-as-a-service, not a managed service. It is an opinionated stack of integrated tools, data pipelines, and governance policies that your organization owns and operates, composed from open-source components, cloud primitives, and proprietary hardening specific to your data and risk profile. The VOS is five layers: **Unified Data and Retrieval Layer.** One enterprise data lake or lakehouse, versioned and queryable, with a single source of truth for embeddings, metadata, and refresh cadence. Every project retrieves from the same place. When you retrain embeddings or update a knowledge base, every application benefits automatically. **Model-Agnostic Inference Layer with Guardrails.** A serving layer that abstracts the model behind a standard interface, whether it's an open-source model on your infrastructure, a fine-tuned proprietary model, or an external API. Guardrails execute before the model runs: input validation, prompt injection detection, PII scrubbing. They execute after: output filtering, tone checking, factuality verification against your knowledge base. **Evaluation and Observability Harness.** A shared evaluation suite that runs against every model change, every deployment, and every production system in continuous operation. Define once; measure everywhere. Production observability feeds back into the eval loop so you know when models drift and can make decisions about retraining or rollback in minutes, not quarters. **Agent Orchestration and Tool-Use Layer.** A substrate for composing multi-step workflows, retrieve from the knowledge base, call an external API, run a calculation, format a response, without writing orchestration logic for each new use case. Tool definitions are registered once and available to every agent. **Deployment and MLOps Pipeline.** Automated promotion from experimentation to staging to production, with feature flags, A/B testing, and instant rollback. A model trained Monday afternoon is in production Tuesday morning if the eval harness approves it. Build this once. Every new project inherits all five layers and can deploy to production in days. ## What Velocity Looks Like in Practice Line-X manufactures protective coatings and sprayed-in bed liners. Before Xivic's intervention, estimating a custom job required a 20-minute phone call, a site visit, and three days of back-and-forth email with a human estimator. The sales cycle was nine weeks long. Xivic built an AI estimator that takes a photo, infers the truck bed dimensions, predicts material costs, and returns a price quote in real time. The MVP took three months to build, primarily because the organization had no shared infrastructure, no eval framework, no guardrails. Crew members built the retrieval layer from scratch. The team wrote custom code to handle estimation edge cases. Deployment was manual and fragile. The system worked. It compressed the estimation pipeline from three days to five minutes. Sales cycle dropped from nine weeks to three and a half. In year one, sales increased 90%, driven entirely by the speed and confidence the tool provided. With a Velocity Operating System in place, that same project would have shipped in weeks, not months. The retrieval layer would already exist. The evaluation framework would validate model output against historical estimates. Guardrails would prevent hallucinated prices. The deployment pipeline would be automated. The team would spend three weeks innovating on the specific value, the estimation logic, the photo-to-dimension model, the material cost predictor, and zero weeks on engineering commodity. Aprilaire, a manufacturer of indoor air quality products, faced a similar inflection point. The organization moved from an on-premises CRM to a cloud data lake. That data lake became the foundation for a direct-to-consumer platform, launched in parallel with internal supply chain optimization. Within the first year, the DTC platform generated over $2M in new revenue. But the velocity payoff was even larger: because the data substrate was shared, cross-functional teams could move fast. By year two, the ROI calculation was not just "revenue from the new channel" but "revenue from every application built on top of this layer, every query that runs faster, every data artifact that doesn't require re-engineering." The compounding is critical. The first project pays for the substrate. Every project after that costs dramatically less. ## What It's Not The Velocity Operating System is not a single platform. It is not "we'll use [vendor offering] and let them handle it." Vendor platforms are useful, but they are not your substrate. They cannot see your data. They cannot enforce your specific guardrails. They cannot run your proprietary evals. When you hand the problem to a vendor, you trade short-term speed for long-term lock-in and loss of institutional knowledge. The VOS is composed. Your data lake lives in your cloud account, your eval harness runs on your infrastructure, your inference layer speaks your company's language and understands your risk profile. A vendor can provide pieces, a managed model API, a hosted vector database, but the glue, the governance, the versioning, the observability: those are yours to build and own. ## Getting Started in 30 Days A CTO can build early momentum toward a VOS in a single month: **Consolidate vector stores.** If you have two or more RAG projects, they almost certainly have separate embedding pipelines and separate vector databases. Pick one, standardize it, migrate. One knowledge base, one embedding model, one versioning story. This single move unblocks every downstream project. **Stand up a shared eval harness.** Define the metrics that matter for your use cases, accuracy against ground truth, latency, cost, safety, and write evaluation code that runs offline. The harness starts small: five test cases per metric. It grows. But the first team to commit to shared evaluation breaks the coin-flip dynamic. **Write model cards.** For every model in production or in development, write a one-page model card that describes its training data, its intended use cases, its known limitations, and its eval scores. These are artifacts that the organization maintains, not individual project teams. Model cards are how you ensure that the third project learns from the first two. **Pick one cross-project guardrail policy.** Don't try to solve guardrails holistically. Pick the one that matters most for your risk profile, whether that's PII scrubbing, injection detection, or hallucination filtering, and build it once, in a shared service. Deploy it in front of every model. Watch it catch problems in seconds that would have taken weeks to debug. ## Why This Compounds The conventional math on enterprise AI is brutal. Project one costs $2M and takes eighteen months. Project two costs $1.8M, you learned things the first time, and takes fourteen months. Project three costs $1.6M and takes twelve months. The slope is slow. With a Velocity Operating System, the math inverts. Project one costs $2M (you're building the substrate). Project two costs $400K and takes two months. Project three costs $300K and takes six weeks. Project four costs $200K and takes three weeks. The substrate is a fixed cost. Every new project is pure application layer. The unit economics improve exponentially. More importantly, the velocity compounds psychologically and organizationally. After the fifth project deploys in three weeks, the organization stops thinking of AI as a capital-intensive, multi-quarter commitment. It becomes operational. Teams in the business start building their own features. The technology moves from the CTO's roadmap to standard practice. That shift, from "how do we build AI?" to "how do we decide what AI to build?", is the difference between velocity as a project management trick and velocity as a compounding organizational asset. Talk to Xivic about standing up your Velocity Operating System. --- # Compound Value: Why One-Off Projects Destroy Multiple Source: https://www.xivic.com/xivic_insight_compound_value.html > Compound Value: Why One-Off Projects Destroy Multiple | Xivic Insights | insights and frameworks from Xivic. *The single-deliverable engagement is the quiet killer of exit multiple. Here's what compounds instead.* ## Hook: The Sprawl Reality An operating partner at a Tier 1 firm reviews 14 months of portfolio spend on digital and AI work. She finds 11 separate vendors, each hired independently. Four different CRM implementations-none speaking to each other. Two abandoned AI pilots that left no playbook behind. A website redesign that didn't integrate with the underlying data model. Search optimization work at three brands, each running its own campaign, each measuring success differently. No shared playbook. No durable asset. No portability. She asks her team: what happens when we exit? What stays? What compounds? The answer comes back quiet: nothing. Each vendor will be replaced. Each project will need to be rebuilt by the next owner. Every brand will start again from scratch. Fourteen months of spend has produced no multiplier. This is the silent killer of exit multiple. Not bad execution. Not technology failure. Something subtler: the architecture of how the work gets bought. ## The Multiple Destruction Pattern Most portfolio AI and digital work arrives as discrete projects. A website redesign, booked as a line item. A CRM implementation, treated as an expense and depreciated. A marketing campaign. An AI pilot. Each engagement has a contract, a deliverable, an end date. On paper, that seems clean. In economics, it's catastrophic. These one-off engagements destroy exit multiple in five ways. **Sprawl:** Each project brings a new vendor, a new tech stack, a new operating model. By year two, the portfolio is fragmented across incompatible systems-a real cost to any acquirer. **Re-work:** Because there's no shared substrate, the next engagement rebuilds the foundations. CRM work doesn't extend the data model. Marketing optimization doesn't feed intelligence back into operations. Search campaigns don't inform pricing strategy. The wheel gets invented repeatedly. **Vendor lock-in:** When each project is discrete, switching costs are invisible until exit, when a buyer inherits technical debt and proprietary integrations. **Playbook loss:** Operating models live in contractor memory. When the engagement ends, the knowledge leaves. The next brand to face the same problem starts at zero. **Measurement amnesia:** What worked in quarter two gets forgotten by quarter four. Results don't compound; they evaporate. These aren't failures of execution. They're failures of architecture. They're the cost of buying projects instead of installing capabilities. ## What Compound Value Actually Is Compound value is an asset that is more valuable in Q4 than Q1, and more valuable after the engagement ends than during it. It doesn't depreciate when the vendor leaves. It doesn't require re-invention at the next brand. It gets stronger as it travels. Three properties define compound value: **Durability** means the asset functions at production grade and doesn't require constant optimization or rework. It's not a campaign that needs refreshing every quarter. It's an operating model that runs. **Portability** means it can move from one brand to another, one geography to another, one use case to another without architectural redesign. A CRM can absorb a new acquisition. A data model can serve new channels. An operating playbook can scale across 17 franchise brands. **Platform effect** means each new application makes the asset more valuable. The second brand using a playbook doesn't pay the discovery cost of the first. The fifth data asset feeding a model makes the model more predictive, not less. These three properties compound exit multiple. Durability means lower post-close remediation. Portability means faster integration. Platform effect means the buyer inherits a system that gets stronger each quarter, not one that depends on vendor relationships. ## The Three Compound Vectors Compound value lives in three places inside a portfolio: **Compound data** means data assets that accumulate and appreciate. Not reports. Not dashboards. Assets: unified customer records across channels, product performance baselines that predict demand, pricing algorithms that improve with volume. Aprilaire built a D2C data lake from customer transaction history and cloud migration, $2M+ revenue in year one, plus 19% cost reduction. That asset didn't end when the project ended. It compounds every quarter as customer cohorts grow and historical patterns sharpen. **Compound systems** are reusable engineering substrates that accelerate delivery and reduce re-work. At Xivic, we call this VOS, the Value Operating System. Instead of building a bespoke CRM integration for each brand, a VOS abstracts the common layer: data intake, transformation, activation. A new acquisition doesn't wait for custom engineering. It plugs into the substrate. The substrate itself improves as patterns emerge across the portfolio. Re-work evaporates. **Compound playbooks** are operating models that travel. They don't live in a vendor's PowerPoint or a consultant's checklist. They're production processes: here is how we run search optimization across 17 franchise brands on a unified platform. Here is the cadence for demand planning. Here is the customer segmentation logic and where it feeds into pricing. When a new brand joins the portfolio, it doesn't hire consultants to figure out demand planning. It inherits a playbook. When the playbook improves, because market data got better, or a new signal emerged, all brands improve at once. ## The Case: The Operating Playbook That Travels Roark Capital's Driven Brands owns 17+ franchise operations: Meineke, Maaco, Take 5 Oil Change, CARSTAR, Fix Auto, and others. Each had separate technology stacks, separate marketing approaches, separate operating cadences. When Xivic unified them on a shared digital platform and implemented search optimization across the portfolio, that wasn't a project. It was the installation of a portable operating playbook. What made it compound: The playbook, "here's how we run demand generation and customer acquisition across franchise networks", doesn't expire when the engagement ends. It doesn't belong to Xivic. It belongs to the platform. When Driven Brands acquires a new franchise, it doesn't hire a new agency. It activates the playbook. When market conditions shift, the playbook improves, and all 17+ brands improve at once. The vendor becomes a non-requirement. The asset becomes a durable part of the platform's operating model. For a PE buyer evaluating the Driven Brands platform, that operating playbook is a material component of value. It means faster integration of future acquisitions. It means lower customer acquisition costs across the portfolio. It means pricing power at exit: a buyer is not paying for past campaign results. A buyer is paying for the portable capability to run acquisition and marketing at scale across a multi-brand franchise network. ## Exit Math: How Compound Value Translates to Multiple Portfolio-level project spend typically commands single-digit multiples relative to revenue. A $2M marketing campaign that drove $15M revenue looks like a 1x leverage ratio to cost. Strategic buyers price it as past work, not future capability. Platform-level operating capabilities that are durable, portable, and produce system-wide effects typically command 2-4x the multiple. A strategic buyer doesn't price it on "what did this cost last year." A buyer prices it on durability: can we rely on this system to work the same way at our company? A buyer prices portability: can we absorb acquisitions without re-work? A buyer prices platform effect: does this get stronger as we scale it? This is where Compound Value Model thinking changes the math. A $2M investment in a playbook that ports across 17 brands, that improves with scale, that doesn't depend on a vendor relationship, is not comparable to a $2M campaign. It's a platform asset. It's where buyer sophistication enters the valuation. ## What This Demands of Portfolio Leadership This shift doesn't happen by accident. It demands a change in how portfolio leadership procures digital and AI work. Not "buy a project." Install a capability. Ask: what asset stays when the vendor leaves? What gets stronger in Q4 than Q1? What travels to the next acquisition? What improves the next application instead of requiring re-invention? The operating partners who ask these questions early, who architect their portfolios around compound capabilities instead of project deliverables, will command multiples their competitors cannot. They'll exit into buyers who recognize that operating models, not vendors, drive value. **Talk to Xivic about installing a compound-value operating model inside your portfolio.** --- # Beyond the Agency Model: How AI Is Rewriting Services Economics Source: https://www.xivic.com/xivic_insight_beyond_agency_model.html > Beyond the Agency Model: How AI Is Rewriting Services Economics | Xivic Insights | insights and frameworks from Xivic. *The billable-hour services firm is dying. What replaces it compounds value instead of billing for it.* ## Hook: Two Proposals It's Tuesday morning in a PortCo CEO's conference room. On the table sit two proposals. The first one is thick-a bound deck from a top-tier firm with a familiar structure. Six-month engagement. Discovery phase, strategy phase, design phase, engineering phase. Sixty pages. Budget: $2.8M across 1,800 billable hours distributed across six tiers of human expertise. Delivery: a website, a roadmap, quarterly reviews. The second proposal is twelve pages. Five weeks. Outcome-priced. Includes a deployed platform, a reusable playbook, and quarterly performance measurement baked in. Budget: $280K. The CEO looks up. Which one looks like the future? For twenty years, the answer would have been clear. The thick proposal represented quality, rigor, and the kind of institutional heft that justifies premium pricing. But something has shifted. The thick proposal assumes that value lives in the hours logged and the layers of review. The thin proposal assumes something different: that value lives in what gets built, what stays deployed, and what gets reused. In 2026, the second proposal is no longer a novelty. It's the model. And the firms that built their economics around the first one are running out of time. ## Why the Old Economics Worked The traditional services model was elegant in its simplicity. Large clients had large problems. Those problems required expensive human hours-strategy that cost $500 per hour, design that cost $300, engineering that cost $200. The firm captured value by stacking those layers: charge for discovery, charge for strategy, charge for design, charge for engineering, charge for management. Every phase created a new invoice line. Handoffs between teams became billable hours. Revisions became change orders. The longer the engagement, the more revenue. The number of junior people you could layer underneath expensive seniors determined your margin. This model worked because it reflected an economic truth: expertise was scarce, and scarce things cost money. If you wanted to apply senior strategy thinking to your business problem, you had to buy the senior strategist's time. There was no alternative. Clients paid for hours because hours were the only unit of value that mattered. And the math was simple: the more people you employed, the more hours you could bill, the more money you made. ## What AI Agents Are Quietly Replacing But the economic truth has shifted. AI agents are doing the work that human hours used to. And they're doing it so fast, and so well, that the hourly model is starting to look like a tax on competence. Content production used to justify a team of writers and strategists. Now an AI system can generate fifty variations of email copy, landing page text, and social media narratives in minutes, not to human-publishable quality yet, but to draft quality that a single strategist can refine in an afternoon instead of eight billable hours. Routine analytics used to demand a data analyst. Now a natural-language query engine can scan your entire dataset and surface patterns faster than a human could set up the SQL. Wireframe generation, first-draft code, basic conversion rate optimization testing, routine SEO audits, first-pass market research, these are all functions that used to justify billable staff. They still need human judgment and interpretation. But they no longer need human grunt work. The honest truth that most large services firms won't say aloud is that the first 70% of many projects, the work that historically justified 40% of the budget, is now machine-doable. The remaining 30% of the work, which used to justify 60% of the budget, is where the real value lives: strategy, judgment, stakeholder navigation, and outcome responsibility. But if a firm is still billing for the first 70%, they're billing clients for the automation of tasks, not the application of expertise. That's not sustainable. It's not even honest. ## The New Economics The firms that see this shift clearly are moving to a different model. They're moving away from time-and-materials and toward outcome pricing. They're pricing engagements on what gets built and what stays deployed, not on the hours it takes to build it. And they're building something that old-model firms rarely talk about: reusable substrates. A substrate is a playbook, a platform, a process, or a dataset that gets built once and reused across clients. It's the diagnostic engine that learns from each engagement and gets smarter. It's the engineering scaffold that scales from a $300K engagement to a $3M transformation without quadrupling the delivery cost. It's the measurement system that stays live after the project ends, generating data that compounds in value. These substrates create a different kind of firm economics: the first client to use the diagnostic substrate pays for its development. The second client benefits from its refinement. The tenth client pays almost nothing for its deployment because the intellectual property is already built. This requires three layers. The first is diagnostic: frameworks, models, and tools that reveal where value lives in a business, the revenue drivers, the cost structures, the hidden leverage points. The second is engineering: the systems, platforms, and automations that turn diagnostic insight into live value. The third is measurement: the operational systems that stay deployed after the project ends, tracking outcomes and feeding learning back into the playbook. Most services firms specialize in one of these layers. The ones that will survive specialize in all three, and they chain them together so that each engagement makes the next one cheaper and better. ## How Buyers Should Choose Firms Today If you're a PE operating partner or a PortCo CEO evaluating services firms in 2026, the old criteria don't work anymore. Industry credentials don't guarantee delivery speed. Team size doesn't guarantee quality. Familiar faces don't guarantee outcome responsibility. You need new questions. - **Show me your reusable operating substrate.** What diagnostic, engineering, or measurement system have you built that carries forward from client to client? Is it proprietary? Has it gotten measurably better with each engagement? If you can't name it, the firm is still selling hours. - **What's your outcome-pricing model?** If the firm quotes only in hourly rates or phases, they're still betting on length, not impact. Ask them to price your specific outcome. If they can't, ask why. - **Show me the operating playbook you'll leave behind.** What instructions, systems, or automations stay deployed after the engagement ends? What's documented and transferable? If everything walks out the door with the consultants, the firm has under-invested in your long-term value. - **What percentage of your delivery is AI-augmented?** This isn't a trick question. It's a disclosure question. If the answer is "none," the firm is behind. If the answer is "we don't measure it," the firm doesn't understand its own cost structure. You want partners who are transparent about where machines amplify human judgment. - **How is my outcome tracked after the contract ends?** Firms that believe in their work keep the measurement engine running. They share dashboards. They tie their own reputation to live performance, not just delivery. Firms that don't do this are optimizing for project closure, not portfolio health. ## What This Means for Your Portfolio If you're buying AI transformation services from an old-model firm, you're overpaying by a factor of three to five. You're also undershooting the outcome by a similar margin. Here's why: old-model firms have to keep projects long to hit their margin targets. They have to layer teams to justify the contract value. They have to scope changes as change orders rather than substrate improvements. And they can't afford to leave measurement systems deployed after the contract ends because they lose the ongoing advisory relationship. The result is that you get a platform, not a system. You get a recommendation, not an operating playbook. You get a final report, not live intelligence. And when the next problem emerges, you call them back, which was the plan all along. The new-model firms price to win on outcome, not to maximize hours. They want to compress the delivery timeline because that's how they scale their playbook. They want to automate and systematize because that's how they protect margin. And they want to leave you with a self-driving measurement engine because that's how they build long-term reputation. It costs you less. It delivers more. And it leaves you with something that compounds. ## What's Next The services industry is halfway through a reckoning. The shift from time-based to outcome-based pricing isn't coming. It's here. The question isn't whether your next services partner will operate on the new model. The question is whether you'll recognize it when you see it, and whether you'll have the clarity to choose it. See what a post-agency engagement looks like, talk to Xivic. --- # From Digital Transformation to AI Transformation: A Playbook for Portfolio Companies Source: https://www.xivic.com/xivic_insight_digital_to_ai_transformation.html > From Digital Transformation to AI Transformation: A Playbook for Portfolio Companies | Xivic Insights | insights and frameworks from Xivic. *Digital transformation gave your portfolio a website. AI transformation gives it an operating model.* ## The Question Your PE Partner Just Asked (and You Don't Have an Answer For) A PortCo CEO walks into her board meeting with a three-year roadmap. The operating partner leans back and says, "Good CRM architecture. And what's the AI layer on top of this?" She pauses. She's built a new sales platform. She's migrated to the cloud. She's got dashboards. But "AI layer" wasn't in the brief, and honestly, it feels like the partner is asking for something that doesn't quite fit on top of the work she's already planned. That moment-the gap between what you've delivered and what the market now expects-is the start of AI transformation. But it's not a layer on top of digital. It's a different operating model underneath. ## Digital Transformation Was About Interfaces. AI Transformation Is About Autonomy. Three years ago, digital transformation meant moving analog work to digital surfaces. You took paper orders and put them in a CRM. You took phone customer service and moved it to a web portal. You took on-premise software and lifted it to the cloud. The work remained the same, the interface changed. The decision-maker was still human. The speed gain came from distribution and access. AI transformation is different in kind. You're not changing the interface. You're changing what makes the decision in the first place. Instead of a human qualified lead, an agent qualifies it. Instead of a human flagging an anomaly in the ledger, a model catches it before the finance team sees it. Instead of a customer service representative deciding if a return is valid, an autonomous system evaluates it and acts. This is not a semantic shift. It's a structural change in the locus of work. Digital transformation distributed work. AI transformation delegates it. ## The Four Layers of an AI-Native Operating Model An AI-native operating model sits on four layers, each with its own diagnostic, build, and measure. **Layer 1: Data.** A single source of truth, structured in a way that models can learn from it. Not a data warehouse. A data lake with model-ready features. The diagnostic here is simple: can your system answer "why?" when the model makes a decision? If you can't trace the data lineage, you don't have a data layer yet. **Layer 2: Models.** Selected, evaluated, and guardrailed models for each decision surface in your business. Not one model. Many. Not off-the-shelf. Validated against your specific decision friction. The measure here is decision quality per dollar spent, accuracy, latency, and cost per decision made. **Layer 3: Agents.** Autonomous workflows that act, not just suggest. A model that recommends a price is still a human decision. An agent that sets the price and monitors for drift is autonomous. This layer is where AI stops being an insight tool and becomes an operating tool. **Layer 4: Outcomes.** Measurement of the compound value each layer creates. Not just lift in a single metric. The loop: better data feeds better models, better models power better agents, better agents create better outcomes, better outcomes generate better data. This is where most portfolios fail, they measure the first win and stop. ## The 12-Month Roadmap **Months 0-3: Diagnose.** Run a Value Friction Inventory (VFI) scan. Interview your operations team. Ask: where do humans spend time making judgment calls? Where would faster decisions change outcomes? Where do you wait for data? You should surface three to five high-value friction points. Pick the three highest-impact ones for the next phase. In month three, you'll have a VFI summary, a prioritized list, and a baseline on today's decision speed and cost. **Months 3-6: Foundation.** Build the infrastructure you'll need to move fast on use cases. Consolidate data from your operational systems into a reusable substrate. We call this the Value Operating System (VOS), the data layer that sits underneath all your agents. This is not glamorous work. It's foundational. By month six, new use cases should be able to go from concept to model in two weeks, not three months. You'll know you're done when your team can deploy a new decision model without touching infrastructure. **Months 6-9: Deploy.** Ship 2-3 agent-level use cases into production. Lean into the friction points you identified in the VFI. Line-X reduced inquiry-to-estimate time by 90% in three months by building an agent to qualify, estimate, and route jobs. Aprilaire launched a direct-to-consumer model with a data-driven demand engine. These aren't pilots. They're live, generating value, and you're learning how agents actually behave at scale. Measure everything. By month nine, your organization should have muscle memory on what it takes to go live. **Months 9-12: Compound.** Measure the compound value your three use cases created together. Quantify the Customer Value Multiplier (CVM) lift, did better data improve decision quality, which improved agent behavior, which improved outcomes? Install the operating playbook: how do you define a friction point, build a use case, and measure it? Then start the next wave. Driven Brands didn't succeed by building one unified platform. They succeeded by installing a playbook and running it across 17+ franchise brands. Month 12 is where you shift from "we did AI" to "we are an AI-native business." ## What Changes First Three operational areas flip to AI-native before almost anything else in a mid-market portfolio company. **Customer Service and First Response.** Your contact center has humans answering questions they've answered a hundred times before. An agent answers them in two seconds, routes complex issues to a human, and learns from feedback. The win is immediate: first-response time drops 80-95%, cost per response falls, and your humans spend time on judgment calls, not repetition. This is where you build team buy-in for the rest of the roadmap. **Lead Qualification and Sales Enablement.** Salespeople spend time on qualifying conversations that a model can run. An AI agent qualifies inbound leads, scores them, and routes them warm. An autonomous system can also surface next-best-action prompts to your sales team. The win is measurable in pipeline velocity and deal size. Dunn-Edwards saw a 43% conversion rate lift and a 30% dealer engagement increase because they automated the qualification layer and let salespeople focus on negotiation and relationship-building. **Internal Finance and Operations.** Reconciliation, forecasting, and anomaly detection. Your finance team spends days on month-end closes because someone has to match the invoices to the receipts. An agent does it. Forecasting happens in real time instead of quarterly. Anomalies are flagged before they compound. The win compounds: better data for planning, faster decisions, less firefighting. ## What to Avoid Most portfolio companies stumble because they mistake the problem they're solving. Don't buy a horizontal AI platform and call it transformation. You'll get a tool. You won't get an operating model. An AI platform is like a cloud infrastructure, it's enablement, not strategy. You still need to know which decisions to automate, what data to use, and how to measure value. The consultancy that sells you platform and walks away is selling you debt. Don't run AI as a side initiative. If the CFO isn't in the room when you plan your data layer, your data layer will be wrong. If the operating partner isn't aligned on what a "win" looks like for each use case, you'll build the wrong thing. AI-native isn't a technology project. It's an operating model change. Treat it like one. Don't procure from old-model agencies. You need a partner who has actually run this playbook inside the portfolio companies you own, not theorists who've read about it. Look for someone who's shipped 2-3 use cases at scale, measured the CVM lift, and lived through the moment when the team realized the agent made a decision they wouldn't have made. That experience matters. And don't skip the VFI diagnostic. Every PortCo has different friction points. The CEO who spends three weeks understanding hers will move faster and waste less than the one who picks someone else's problem to solve. ## The Operators Who Win The CEOs who win the next 24 months won't be the ones with the best AI models. They'll be the ones with the best operating playbook, the repeatable system for identifying friction, building agents, measuring value, and doing it again. Start your AI transformation diagnostic with Xivic. --- # The Agentic Factory: How Autonomous AI Is Rewiring Industrial Operations Source: https://www.xivic.com/xivic_insight_agentic_factory.html > The Agentic Factory: How Autonomous AI Is Rewiring Industrial Operations | Xivic Insights | insights and frameworks from Xivic. *Manufacturing is becoming agentic. The only real question is whether your company or your distribution channel owns the autonomy layer.* ## The Distributor's Dilemma Picture this: Your largest distributor deployed an AI agent last quarter. It takes inbound customer calls, routes orders, recommends your products based on use case and inventory, and closes the loop inside their system-all without a human hand on the process. Every conversation your brand used to own now happens inside your distributor's agent. The agent has learned the product specs, the margin architecture, and the customer profiles. In six months, it will know your customer base better than your own team. What does your margin look like in three years? Your brand data? Your direct customer relationship? This isn't speculative. It's happening now in industrial distribution, and it will accelerate. The manufacturers who move first to build the agentic layer inside their own operations will own the relationship, the data, and the margin. The ones who don't will become interchangeable suppliers behind someone else's agent. ## What Agentic AI Actually Is Chat AI answers questions. Agentic AI acts. It retrieves information, decides what to do, calls tools, writes to systems, and closes loops-all in sequence, all autonomously. An agent doesn't wait for a human to interpret a question and type a response. It sees an incoming request and executes a workflow. Here's a concrete example: A customer submits an inquiry through your website asking for a custom product configuration and estimate for a fleet application. A chat AI can draft a nice response explaining what you need. An agentic AI does this instead: it pulls the product specification matrix, analyzes the customer's use case against the rules, generates a custom engineering recommendation, computes the pricing (including volume discount, application-specific surcharges, and margin targets), writes the quotation to your CRM, books a dealer appointment on the salesperson's calendar, sends a confirmation email to the customer, and logs the interaction for your forecasting engine, all in under 60 seconds, with zero human touch. That's the agent layer. The gap between chat and agent is the difference between information and execution. Agents are dangerous precisely because they're useful. They can make decisions at scale. Manufacturers who control that decision-making layer own the future relationship. Those who cede it, to a distributor, a marketplace, a retailer, become suppliers to someone else's agent. ## Why Manufacturing Is the Next Frontier Manufacturing has unique structural advantages that make it the highest-value frontier for agentic AI. Five dynamics converge. First, manufacturing involves complex product configuration and rule-based decision-making at scale. An HVAC system must match the building's load, climate zone, duct layout, and efficiency targets. A coating system must account for substrate, environmental exposure, application method, and performance guarantees. A distributor's sales team memorizes fragments of these rules. An agent masters all of them instantly and applies them consistently to every customer. Configuration is the work that agents were made for. Second, distributor and dealer networks are expensive, high-friction, and slow. A customer question arrives at a dealer. The dealer calls the manufacturer rep. The rep emails engineering. Engineering responds. The rep calls back the dealer. The dealer calls the customer. Four days pass. An agentic system compresses this to four seconds. The friction is so obvious, and the cost of the network so high, that displacement is inevitable. Third, after-sales service and parts logistics consume disproportionate margin and customer goodwill. Preventive maintenance schedules, parts availability checks, service call booking, warranty claim routing, these are agent workflows waiting to happen. A manufacturer's service network can serve customers in real time instead of handing them off to a call center, a portal, or a field rep with a clipboard. Fourth, modern manufacturers have accumulated rich product and usage data, field telemetry, application performance, customer profiles, seasonal patterns. This data is exactly what makes agent decision-making accurate and trustworthy. The data layer was built for this. Agents are the output layer that justifies the investment. Fifth, direct-to-consumer pressure is mounting. E-commerce taught customers to buy frictionlessly. B2B manufacturing still requires a distributor, a sales call, a specification sheet, a quote, and a negotiation. Agentic commerce collapses that workflow. A DTC manufacturer that can configure, quote, and sell through an agent becomes structurally harder to disintermediate. The dealer can no longer claim exclusivity or personal touch, the agent beats the dealer on speed and consistency. ## The Four Agentic Surfaces Inside a Manufacturer Agents don't deploy everywhere at once. They surface in four distinct operational domains. Each has different economics and deployment order. **Procurement & Supply Chain Agents.** These forecast demand from customer signals, predict supplier lead times, negotiate replenishment orders, and optimize inventory turns. They work with upstream data and internal systems, lower external friction, higher leverage on working capital. These often come first because the ROI is internal and measurable. **Operations & Maintenance Agents.** These predict equipment failures before they occur, inspect quality in real time, detect process anomalies, and recommend corrective actions. They connect to sensor data, production schedules, and maintenance workflows. The upside is availability and yield improvement. These follow once the data foundation is solid. **Dealer, Distributor & Partner Agents.** These manage partner onboarding, co-op marketing spend, training delivery, and performance analytics. They're the channel interface, replacing quarterly business reviews and email with real-time agent-mediated partner engagement. These unlock margin and competitive lock-in because they make dealer networks more productive without replacing them. **Customer & Commerce Agents.** These configure products, generate estimates, schedule service, support DTC sales, and manage customer inquiries. They're the revenue-facing layer. Estimates, configuration, and booking are the first high-volume workflows that shift to agentic. These are where the customer relationship lives. ## Proof in Production Three manufacturing partners prove the foundation and the path forward. **Aprilaire ($740M residential HVAC):** Moved DTC on its website and reached $2M+ in Year 1 DTC revenue with a 42% conversion lift over traditional channel sales. The critical enabler was the data layer: product specifications, application rules, customer usage patterns, and seasonal demand. That data layer, not yet agentic, but rule-based, became the substrate. When you add autonomous configuration and quotation on top, you get the agent surface. The cost reduction from migrating to cloud infrastructure (19% reduction in data and processing costs) freed up margin to invest in the autonomy layer. **Line-X ($2.5B protective coatings and franchise):** Built a mobile-first specification engine that reduced the inquiry-to-estimate workflow from days to hours. The sales cycle compressed from three weeks to 1.5 weeks. The MVP shipped in three months. This is proto-agentic work: rules, integrations, and decision-making, but not yet fully autonomous. The next evolution, fully autonomous agent that takes a photo, evaluates substrate and geometry, generates the spec, books the spray appointment, and routes to the franchise partner, is the natural step forward. **Dunn-Edwards Paints ($880M B2B and DTC):** Achieved 43% conversion rate lift on the e-commerce site and 30% dealer and partner engagement lift through a unified digital platform. The dealer network became more productive because the agent-ready data on colors, specifications, and regulatory compliance moved onto a shared platform. The next layer is an agent that recommends color palettes based on project type, tracks inventory across the network, and proactively alerts dealers to inventory gaps and seasonal demand shifts. **Driven Brands (17+ franchise brands, Roark Capital portfolio, $1.8B+ combined revenue):** Unified the digital footprint across franchised brands (Servpro, Anixter Systems, Coverking) with a cross-portfolio search and recommendation layer. The platform allows a customer to find services across the portfolio. Agents can now route customer requests to the best brand or franchise partner in the network, manage co-op marketing spend across franchises, and ensure brand compliance while maximizing fill rates. This is the playbook for compound scaling: one data layer, one agent architecture, deployed across a portfolio of brands. Each of these manufacturers now has the foundation on which agentic layers naturally sit. The data is clean, the rules are codified, and the integrations are in place. Adding autonomous agents on top is a 90-day sprint, not a two-year transformation. ## The Three-Year Arc By 2028, agentic surfaces will have moved from edge case to operational baseline in manufacturing. Here's what that trajectory looks like. Agent-mediated commerce becomes the default. Customers configure, quote, and purchase without touching a salesperson. Dealers and distributors still exist, but they're now specialized service layers, not transaction bottlenecks. The agent handles the commodity trade; the human handles the relationship and the problem-solving. Manufacturers who built this layer internally will have a direct customer feedback loop that their competitors will envy. Agent-managed dealer networks become a competitive moat. The manufacturers whose agents actively manage partner onboarding, co-op spend, training, and performance, not passively reporting on it, will have dealer networks that execute faster and more consistently than their peers. Dealer satisfaction will improve because the agent removes friction, not because the agent replaces the dealer. Agent-driven predictive operations shift margin structurally. Manufacturers who own the maintenance and reliability layer through agentic predictive systems will shift from reactive service models (customer breaks something, pays for emergency service) to proactive models (agent predicts failure, customer buys preventive maintenance). That's a margin story. It's also a customer satisfaction story. The manufacturers who move first will be structurally harder to disintermediate. A distributor can't as easily claim they own the customer relationship if the manufacturer's agent is already on the call, in the configuration, and in the recommendation. The autonomy layer is the new barrier to entry. ## How to Start Deploying agentic AI at scale doesn't require a multi-year program. The manufacturers moving fastest are following a tight four-step approach. **Pick one agentic surface as your first bet.** Don't boil the ocean. If your highest friction is in dealer engagement, start there. If it's in after-sales service requests, start there. If it's in DTC commerce, start there. One surface, one bet, one outcome metric. **Run a VFI scan to find the highest-friction point in that surface.** VFI (Velocity, Friction, Impact) maps the workflow. Where is the manual handling? Where is the delay? Where is the repeat work? That's where the agent lives. A 30-day VFI project will show you the exact workflow the agent should automate and the data it needs to succeed. **Build the foundation before the agent.** The foundation is the data layer and the VOS (Vendor Operating System) substrate. Clean data, codified rules, and API-first integrations between your systems. Don't deploy an agent into a chaotic data environment. You'll get chaotic decisions. VOS is the orchestration substrate that connects your core systems (CRM, ERP, inventory, partner systems) so the agent can read and write across them. Typically a 12-week sprint for a manufacturing company. **Ship a scoped agentic use case in 90 days, measure, and compound.** Define the agent narrowly. Configuration for one product line. Estimates for one dealer network. Predictive maintenance alerts for one equipment class. Measure adoption, accuracy, and the specific outcome (speed, cost, margin). Once it works, compound: add a second product line, a second network, a second equipment type. Each win funds the next bet. ## The Autonomy Layer Is the Battleground For the past two decades, manufacturing has been a digital-interface industry. Websites, portals, CRM, e-commerce, all of it designed to inform human decisions. That era is ending. The next era is agentic: autonomous systems that don't inform decisions, they make them. The manufacturers who build the autonomy layer inside their operations own the factory of the future. They own the customer relationship, the data, and the margin. The ones who don't, who cede the autonomy layer to distributors, retailers, or marketplaces, will supply it. That's not a technology prediction. It's a structural one. The time to move is now, while the gap between the first movers and everyone else is still visible. In three years, it will be too late. **Talk to Xivic about building the agentic layer inside your manufacturing operations.** --- # Agentic Manufacturing: Five Plays to Deploy Autonomous AI in Industrial Operations Source: https://www.xivic.com/xivic_insight_agentic_manufacturing.html > Agentic Manufacturing: Five Plays to Deploy Autonomous AI in Industrial Operations | Xivic Insights | insights and frameworks from Xivic. *Concrete moves manufacturers can run in the next 90 days - not theory, not next decade. Ranked by time-to-value and defensibility.* Most manufacturing leaders have absorbed enough "AI will transform your business" messaging to be skeptical. What they need is clarity: which autonomous moves should we actually run, and when? The Agentic Factory concept sets the strategic frame, data, systems, operating model aligned around autonomy. This piece gives you the five plays you can sequence and scope within 90 days, each with a trigger, a clear business case, and a defensible proof point. The five plays below are not theoretical. They're sequenced variants of work already embedded in Xivic's manufacturing portfolio, from demand forecasting to dealer engagement, from self-serve quoting to predictive parts, and service automation at scale. Pick the right first play for your current state, and the subsequent ones compound on your data and organizational momentum. ## Play 1: Autonomous Demand Forecasting **Trigger:** Your forecast is quarterly, manually prepared, and systematically off by 15-25% WAPE. Planners are frustrated. Inventory carry costs are creeping up. Production schedules shift every month because the forecast wasn't there to anchor them. **What it replaces:** The quarterly planning cycle where a human team pulls last year's actuals, adjusts for "what we heard from sales," and publishes a number that the organization treats as gospel for 90 days. The agent replaces this with a continuous ingestion layer, sales velocity, dealer inventory, regional weather signals, macro demand indices, and issues an updated forecast daily. **What it saves:** Industry benchmarks for forecast error reduction from statistical models to ML to agentic typically show 20-40% lower WAPE. That translates to direct inventory carry-cost savings (8-12% inventory reduction is typical), stock-out revenue prevention, and reduced production schedule thrash. Aprilaire's cloud data lake foundation delivered $2M+ DTC revenue impact Year 1 plus 19% operating cost reduction, the agentic demand layer compounds on top of that substrate. **Proof and how to scope it:** Aprilaire anchors the data foundation; the agentic forecast layer is the next evolution. Start with a 90-day diagnostic, map your current forecast inputs, latency, and error sources. Then run a parallel 90-day pilot on one SKU family (high-volume, demand-volatile) and measure WAPE reduction and safety-stock impact against your baseline. ## Play 2: Agentic Dealer & Distributor Portals **Trigger:** Your dealer network is growing or consolidating. Onboarding new dealers or territories takes months. Training is inconsistent. Performance visibility is opaque. You're losing margin to dealer churn, and field teams are burned out by training and compliance work. **What it replaces:** The traditional dealer portal, product catalog, performance dashboards, training materials, co-op claim forms, where dealers log in, find the content they need, and you hope they onboard and stay trained. The agentic layer replaces this with an autonomous agent that onboards dealers, adapts training to their profile, flags performance drift, recommends co-op spend, and escalates only when human judgment is required. **What it saves:** Field-team time (40-60% reduction in onboarding and training hours), training cost, dealer churn reduction. Dunn-Edwards achieved 30% dealer engagement lift on a pre-agentic platform, the autonomous layer pushes that further. Driven Brands' franchise platform (17+ brands unified under Roark Capital) demonstrates the multi-brand scalability, once the agentic substrate is live, you can replicate the experience across franchises with minimal incremental cost. **Proof and how to scope it:** Driven Brands' unified digital platform for franchise operations is your north star. Start with a 60-day foundation engagement: diagnostic on current dealer pain points, system integration mapping (ERP, dealer platform, training systems), and agentic workflow design. Then run a 60-day pilot on a subset of dealers (new cohort or a single region) and measure onboarding time, training completion rates, and engagement lift. ## Play 3: Self-Service Agentic Estimate & Quote **Trigger:** Your customer inquiry-to-quote cycle is measured in weeks, not minutes. Sales teams are bottlenecked. Customers go silent and buy elsewhere. Dealers complain about lead quality and time lag. You're leaving revenue on the table every week. **What it replaces:** The manual estimate process, customer fills out a web form or calls a dealer, request sits in a queue, sales or customer service calculates the estimate, and 5-10 days later you follow up. The agent replaces this with real-time ingestion of customer specs, instant lookup of product pricing and configurations, and an estimate issued in seconds, plus automatic booking of the next step (demo, order, site visit). **What it saves:** Sales cycle length (3 weeks to 1.5 weeks typical), lead loss from slow response, dealer frustration about lead quality and timeliness. Line-X's online estimator delivered 90% inquiry-to-estimate time reduction and halved the sales cycle in a 3-month MVP, and that was rules-based. The agentic version handles configuration complexity, custom requests, and dynamic pricing without human intervention. **Proof and how to scope it:** Line-X is your proof: rules-based estimate automation delivered massive conversion lift and cycle-time compression. The agentic evolution handles the 20% of complex or edge-case quotes that the rules engine had to hand off. Plan a 90-day production ship: weeks 1-4 for requirements and API integration (with ERP, configurator, CRM), weeks 5-8 for agentic workflow development and testing, week 9-12 for launch and iterative improvement based on conversion and escalation metrics. ## Play 4: Predictive Inventory & Parts Agents **Trigger:** Your after-sales parts business is profitable but leaving money on the table. Service truck utilization is at 50-60%, trucks roll out half-loaded. Stock-outs and emergency shipments are expensive and hurt CSAT. Field teams spend time chasing parts instead of servicing. **What it replaces:** Static parts inventory planning (regional hubs stock based on historical averages) and manual service routing (dispatchers assign work, field teams optimize their own route). The agent replaces this with predictive failure signals (part failure patterns, service history, equipment age/usage), dynamic pre-staging of parts at hubs, and autonomous route optimization that considers inventory and likelihood of multi-unit or multi-part jobs. **What it saves:** Truck-roll efficiency (15-30% utilization improvement typical; best-in-class hits 40%+), stock-out revenue loss, warranty cost reduction, and improved CSAT because service is faster and less friction-laden. Aprilaire's cloud foundation and Xivic's data work across the portfolio anchor the prerequisite substrate, predictive parts is the next autonomy layer on top. **Proof and how to scope it:** Predictive maintenance and parts optimization is standard practice in best-in-class manufacturing operations. Start with a 90-day diagnostic on your parts-failure data, service history, and regional inventory positioning. Then run a 90-day pilot on a single service region or parts family, measure truck utilization, stock-out incidents, and cost per job against baseline. Once validated, scale to remaining regions. ## Play 5: DTC Service & Support Agents **Trigger:** Your DTC or B2C support volume has grown faster than your team. CSAT is eroding. Customers are frustrated by response time and repetitive questions. Your team is burned out. **What it replaces:** The traditional support queue, customers submit tickets (chat, email, phone), your team reads them, looks up account/order/product info, and manually responds. Repeat questions about order status, returns, troubleshooting, and warranty take the most time. The agent replaces this with an autonomous service layer that resolves routine issues end-to-end, order status lookup, return initiation, product troubleshooting, warranty-claim routing, and escalates cleanly to humans when judgment or empathy is required. **What it saves:** Cost per ticket (40-70% automation typical on well-grounded agents), response time (minutes vs. hours), escalation rate, and CSAT often improves because resolution is faster. Aprilaire's DTC foundation delivered 42% campaign conversion lift and $2M+ Year 1 revenue impact, the service agent layer compounds on that commerce foundation, handling the post-purchase experience at scale. **Proof and how to scope it:** Aprilaire shows the commerce side; the service agent is the natural follow-on. Scope this as a 90-day engagement: 30 days to wire the agent to your support systems (ticketing platform, order database, product KB), define escalation rules, and build the conversational flows for your top 10-15 support use cases. Then run a 60-day pilot, agent handles a subset of incoming tickets, humans monitor and escalate when needed. Measure automation rate, customer satisfaction, and cost per ticket. ## Sequencing: Which Play to Run First The right sequence depends on your current state, but the general hierarchy is defensibility and time-to-value. **Start with Play 3 (Estimate Agent)** because it drives revenue impact directly, builds customer relationships that are hard to disintermediate, and is fastest to prototype (90 days to production). **Play 2 (Dealer Agent) second**, because it defends margin against channel disintermediation and scales dealer productivity at low incremental cost. **Play 5 (Service Agent) third** for a scalable cost structure that improves CSAT, it compounds on DTC or B2C momentum. **Plays 1 and 4 (Forecasting and Inventory)** come fourth and fifth; they require deeper data foundations and are most valuable once you've built organizational fluency with agentic workflows. These five plays aren't theoretical. They're sequenced moves already embedded in Xivic manufacturing portfolios, Dunn-Edwards, Aprilaire, Line-X, and the Driven Brands franchise platform. The question isn't whether autonomous agents work in manufacturing. It's which one you run first, and how you sequence the next four to compound defensibility and value. **Talk to Xivic about sequencing these plays for your manufacturing operations.** --- # The CTO's Dilemma Source: https://www.xivic.com/xivic_insight_ctos_dilemma.html > The CTO's Dilemma | Xivic Insights | insights and frameworks from Xivic. You know the meeting. It's Tuesday at 2 PM. Your VP of Engineering-sharp, ambitious, hired six months ago from a Series B-pulls up a slide deck titled "ERP Modernization 2026." The architecture is beautiful. Microservices. Event-driven. Cloud-native. Kubernetes. The team that built it is hungry. The timeline: 24 months. The budget: $12M. The implicit assumption: we start fresh, or we die. Meanwhile, your CFO is staring at the slide with the number that matters: the 17-year-old ERP system on that slide is processing 40% of your company's revenue. It's running on AIX. It's written in COBOL and custom C. No one under 40 understands it. But it hasn't failed in four years. Your board just approved a $40M commitment to AI. They want agents, they want to move fast, and they absolutely do not want a rebuild that bleeds into 2027. This is the CTO's dilemma of this decade. It isn't new, the question of "rebuild or refactor" has been asked since the 1980s. What **is** new is that for the first time, you have a third option that actually works. And it might save your company millions and six months you don't have. ## Why the Modernization Playbook Usually Fails Let's start with honesty. The textbook has three plays for legacy modernization: **Rip-and-replace** makes CFOs wince. You turn off the old system and light up the new one. The appeal is total: clean break, no technical debt, full cloud. The reality: according to McKinsey, 60-70% of enterprise ERP replacements run over budget. Gartner found that 25-30% of large-scale re-platforming efforts get terminated mid-flight. That's not a failure rate, that's the expected outcome. And that's the *successful* 70%. The failed ones tend to stay quiet. **The strangler pattern** is the "smart" choice. You build the new system next to the old one, gradually migrating data and workload across the boundary. It *sounds* elegant. In practice, the strangler fails when the boundary is fuzzy, which it always is. The legacy ERP doesn't just process orders; it manages inventory, triggers compliance events, feeds the general ledger, and holds five years of audit trail that no one has fully documented. You end up with two systems, both running in critical paths, both degraded because each is now responsible for half the workflow. You've traded "legacy debt" for "system integration debt," and the latter is worse. **Lift-and-shift** migrates the old system to the cloud, which is progress, but it's not modernization. You still have the same monolith. You've just moved it from a server room to AWS. The interface is still terrible. The data model is still inflexible. You've spent $2-3M to solve the wrong problem. What all three approaches share is this: they assume the system itself is the problem. So they spend a year and a half solving it. But the real problem, for most businesses, is that the system has become invisible to the rest of your architecture. The workflow layer can't talk to it. The analytics layer can't see inside it. Mobile apps can't consume it. Your AI agents can't act on it. The system isn't broken. The *interface* is. ## The New Option: Agent-Wrapped Legacy Here's what became possible in the last 18 months: you can wrap a legacy system in a purpose-built agent layer that exposes its capability as clean, composable APIs without touching the underlying code. This isn't middleware. It's not a fancy data integration layer. It's a small, focused AI agent, or a cluster of agents, that sits between your legacy system and everything else. Here's how it works: The agent learns the legacy system the way a human operator does. It navigates the UI (text-based, web-based, whatever). It understands the business rules encoded in the system's behavior. It extracts data from reports, processes forms, triggers transactions. And crucially, it enforces a clean API contract at the boundary: structured input, validated output, deterministic behavior. A practical example: imagine a 20-year-old AS/400 order-to-cash system. It's a closed, proprietary system. The only way to interact with it is through green-screen terminal commands that only one engineer remembers. An agent-wrapped modernization looks like this: - You define the API contract: given an order ID, return fulfillment status, expected delivery, real-time inventory. - You build an agent that knows how to navigate the AS/400 terminal, extract the relevant data, parse it, and return it as JSON with a guaranteed schema. - You stand up that agent in a managed runtime with observability, versioning, and governance. - Now your mobile app talks to the agent. Your analytics pipeline talks to the agent. Your AI agents talk to the agent. The AS/400 stays exactly as is. The same pattern works for homegrown ERPs, legacy COBOL systems, even paper-based processes that have been digitized with OCR and LLM extraction. Why does this work now? LLMs have finally reached the capability threshold. They can navigate complex UIs without being explicitly programmed for each menu. They can infer business logic from examples. They can enforce structured output. A year ago, this was 60% reliable. Today, with proper schema enforcement and error handling, it's north of 85%. ## The Economics of Interface vs. Rebuild Let's talk numbers, because this is where the dilemma resolves. A full rebuild of a complex legacy system, full scope, enterprise-grade, with all the edge cases, typically costs $8-25M and takes 24-36 months. The failure rate is real. But let's assume you execute well. You've derisked it. You've got a strong team. You spend $12M and you finish in 28 months. That's a win. Agent-wrapped modernization of the *same system* costs $400K-$2M per system. The timeline is 60-120 days. The success rate is 85% or higher because you're not rewriting business logic; you're wrapping it. Now here's the math that matters: the agent-wrapped version doesn't eliminate the rebuild. It defers it. But in those 60-120 days, you've unblocked your entire organization to act on the legacy system. Your data engineers can build pipelines. Your product team can build new experiences. Your AI teams can compose the legacy system into workflows. And you're generating measurable operating improvements, reduced manual work, faster fulfillment, fewer errors. Those improvements fund the rebuild. They justify it. And they buy you the political runway to do it on the right timeline, not the emergency timeline. The second-order effect is even more powerful: once you've wrapped the system, you understand it differently. The manual work it eliminated, the pain points you surfaced, the data flow you exposed, all of that tells you whether the rebuild is actually necessary, or whether the agent layer is good enough to last another 5-10 years while you modernize something else. ## Building the Agent Layer Right This isn't magic, and it's not a patch. Here's what matters: **Enforce schemas at the boundary.** Don't trust LLM outputs directly. Define the exact structure the agent is required to return. Validate it on every call. If the output doesn't match, the agent retries or fails gracefully. Garbage in, garbage out is an architecture problem, not an LLM problem. **Build governance from day one.** The agent layer can become new legacy just as quickly as the system it wraps. Version it. Define deprecation paths. Track which downstream systems depend on which agent endpoints. Don't let it become a blackbox tangle of chained LLM calls. **Define data contracts.** Just because the agent "figured out" what the legacy system returned doesn't mean you skip the step of documenting it. The agent should expose a data contract: this endpoint returns these fields, in this format, with these guarantees of completeness and timeliness. Update it as the legacy system evolves. **Observe deeply.** You need visibility into every agent interaction with the legacy system. Not just success vs. failure, but latency, retry patterns, edge cases the agent encountered. This observability is how you know when it's time to rebuild the underlying system. **Don't build it for the team that's quitting.** Agent-wrapped modernization only works if you're building a platform that lasts. If the intent is "we'll do this for six months until we find someone who knows the system," you've already lost. Build it for durability. ## The New CTO Playbook: 12 Months Here's a concrete timeline that works: **Q1: Inventory and VFI scan.** You're looking for the highest-friction legacy interfaces, the systems that create the most manual work, slow down the most processes, block the most new initiatives. Not the biggest system. The one that, if it were modern, would unblock the most value. Usually it's order management, subscription billing, or core platform infrastructure. **Q2: Wrap and prove.** Pick two to three critical legacy systems. Build the agent layer. Stand it up in production with real traffic if possible, or in a realistic staging environment. Measure: time to return, reliability, schema violations, operator overhead. Prove the model works for your organization. **Q3: Cross-system workflows.** Extend the agent layer to enable workflows that span multiple legacy systems. This is where the real value surfaces. Customer onboarding that touches three different systems. Inventory management across the warehouse system and the finance system. The manual steps disappear. **Q4: Fund the rebuild.** Use the operating improvements, reduced manual work, faster cycle times, fewer errors, to justify investment in selectively rebuilding the 1-2 systems that genuinely need it. The rebuild is derisked because you understand the system intimately now. And you're not doing it because you *must*; you're doing it because you *can afford to*. This sequence spreads the risk, generates operating improvements along the way, and gives you the data and organizational alignment to make the rebuild decision wisely. ## The Pattern: Xivic's Velocity Operating System What enables this playbook is having the right infrastructure. The agent layer isn't a one-off script; it's a system, shared agent runtime, shared data and event layer, unified policy and permission enforcement, observability across all agents. This is the pattern underlying Xivic's Velocity Operating System: a platform designed to compose agents across your stack, legacy and modern, internal and external, AI-native and retrofit. The VOS is the substrate that makes agent-wrapped modernization sustainable, not tactical. We've built this pattern with multiple PE-backed platform companies that have inherited complex legacy stacks, and the economics are consistent: 70% of the value of a full rebuild, at 15% of the cost, in 10% of the timeline. ## The Anti-Patterns to Avoid Before you start, know what kills this approach: Don't let the agent layer mask bad data. If the legacy system returns inconsistent data, the agent shouldn't paper over it with inference. It should flag it, so you *know* the rebuild is necessary. Don't skip the governance conversation. The first agent works great. By agent fifteen, without versioning and deprecation, you've built a new legacy system inside the old one. Don't assume the agent replaces the rebuild forever. It buys time and derisks the decision. But if the legacy system is a genuine business constraint, slow, unreliable, impossible to audit, the rebuild still happens. The agent layer just means you're not doing it in crisis mode. Don't underestimate the operational load. The agent layer needs monitoring, updates, and support. Budget for it. ## The Honest Answer Here's what I've learned as a CTO who's sat in the "rebuild or refactor?" meeting a hundred times: The textbook says you have to choose: rip-and-replace, strangler pattern, or accept technical debt. The honest answer, for 15 years, was "none of these work well enough, so we patch and pray." The new honest answer is: "We can now modernize the *interface* to the system instead of the system itself. We get 60-80% of the value, at 10-20% of the risk and cost, and we buy the runway to do the rebuild on our own timeline." That changes everything about how you prioritize. It changes how you explain modernization to your board. It changes what you promise your best engineers: not a two-year migration project, but a six-month win that funds the long-term work. The best CTOs of this era aren't the ones who finish the big rebuild first. They're the ones who realized the rebuild wasn't the point. The point is unblocking the business. Sometimes that's a rebuild. Sometimes it's an agent layer that costs a tenth as much and ships in a tenth of the time. Your job is to know the difference. The dilemma isn't gone. But for the first time, you have a choice that actually resolves it. --- # The CMO's AI Stack Source: https://www.xivic.com/xivic_insight_cmo_ai_stack.html > The CMO's AI Stack | Xivic Insights | insights and frameworks from Xivic. The organic search traffic to your competitor's biggest content hub dropped 34% in the last eighteen months. Their paid CAC climbed 47%. Their email open rates are down to 18%. And their CFO is asking why the marketing stack still costs $2.8M annually. You know the feeling. You've been there. What they don't tell you is that the traffic didn't disappear. It moved. Your buyers stopped searching Google for "best CRM for mid-market." They opened ChatGPT and asked an AI. They stopped reading your comparison guides. They asked Claude to summarize your competitors. They stopped clicking your nurture emails. They delegated the first fifteen emails in your drip campaign to a Gmail agent that already sorted them. The marketing automation stack that dominated the last decade-HubSpot, Marketo, Pardot, the whole ecosystem-wasn't built for this world. It was engineered for a buyer who read emails, who engaged with mid-funnel content, who left a trackable breadcrumb trail of clickthroughs and form submissions. That buyer is disappearing. The CMO who hasn't noticed is already a quarter behind. The CMO who's noticed but doesn't know what to do is about to get disrupted. The CMO building the new stack, agentic demand gen, is the one who'll own the next cycle. ## The Data That Should Scare You Let's start with what's actually happening in the market. According to Semrush and similar traffic intelligence platforms, organic search referral traffic to B2B SaaS websites is down 15-25% year-over-year across 40% of tracked categories. But that's not the full story. Direct referral traffic from AI assistants, ChatGPT, Perplexus, Claude, Copilot, Gemini, has grown 10-20x in the last eighteen months. Your brand is being mentioned in AI outputs, but you're not getting the click. Your buyer is getting the answer without ever landing on your site. For email, it's worse. Gmail's Priority Inbox now routes commercial emails with aggressive AI filtering. Third-party email agents that auto-sort, auto-archive, and auto-delete marketing messages are becoming table stakes. One Litmus study found that 32% of mid-market marketing teams saw email engagement drop below 16% open rates last quarter. Below 16%. Paid media isn't immune. Facebook and Google's CAC is compressing across B2B because AI agents are doing the buying, and they don't fall for carousel ads. If your conversion pathway requires three touchpoints and a retargeting sequence, an AI agent will skip it and move to the competitor who offers a live chat or configurator. This isn't a Google algorithm update. This is a buyer-side infrastructure shift. The infrastructure is AI. ## Why Marketing Automation Peaked The marketing automation platform was built on three assumptions. Every one of them is now wrong. **Assumption one: Humans read emails.** Drip campaigns, nurture sequences, and email scoring engines were designed around the idea that your buyer would open an email, scan a headline, and click a link. The entire model, cadence, personalization, subject line testing, assumed human attention. But today, AI agents triage your buyer's inbox. They ask: is this urgent? Is this relevant? Should my human even see this? Ninety percent of your nurture campaign ends up in a folder the human never opens. You optimized for a reader who isn't there. **Assumption two: Engagement signals predict intent.** Marketing automation platforms built scoring models around clicks, email opens, time on page, and form submissions. These signals, the logic went, predicted sales readiness. But now your audience is half-human and half-AI. An AI summarization agent might click every link in your resource center to summarize it. A competitor's bot might hit your site 40 times a day running competitive intelligence. Your MQL scoring model is now 40% noise. A human clicked that email because they were interested. A bot clicked it because it indexed your content. How do you tell the difference? You can't. Your lead scoring is broken. **Assumption three: Humans route and accept leads.** The HubSpot workflow where an MQL gets routed to a rep, the rep accepts or rejects the lead, and a point-of-time conversion decision gets recorded, that assumes a human makes a binary call. But AI SDRs now do the first-pass triage. They read the company description, check the funding stage, scan the persona, and make a routing decision. By the time a human touches it, the AI's already decided. Your lead routing workflow is now an extra step. Your sales team thinks you're sending them bad leads. What you've sent them is leads that an AI already pre-qualified. You don't have data on what AI decided. The marketing automation platform is still firing. But it's firing at a buyer that no longer exists. ## The New Stack: Agentic Demand Gen What replaces the old model? Not tweaking HubSpot. Not AI-powered email send time optimization. Something fundamentally different. Agentic demand gen treats AI as three things simultaneously: your primary audience, your primary channel, and your primary co-marketer. The stack has three pillars. **Pillar one: AI-discoverable content (AEO/GEO).** Your buyer isn't searching "best CRM" on Google anymore. They're asking ChatGPT: "What CRM would you recommend for a Series B SaaS company with 50 employees?" ChatGPT isn't going to link to your comparison guide from 2023. It's going to generate an answer. If your brand appears in that answer, you win the impression. If you don't, you're invisible. This requires a different content strategy. You're not optimizing for keyword volume and backlinks. You're optimizing for being the right source that generative models cite. Answer Engine Optimization (AEO) and Generative Engine Optimization (GEO) are the new disciplines. Your content needs to: - Answer specific, narrow procurement questions (not broad category research) - Exist in formats that AI can easily parse and cite (structured data, Q&A formats, data-rich comparisons) - Build enough authority and citation frequency that AI models weight your brand heavily - Be regularly updated so it remains current in model training data A single piece of content in the old model might get 300 organic clicks a month. In the new model, if ChatGPT cites your brand in 10% of relevant queries, and 50,000 people a month ask that question, you're getting 5,000 "impressions" in an AI context where your competitor gets zero. You never see the click. But you own the mind. **Pillar two: Agentic buyer experiences.** Your website should be an autonomous agent. Not a brochure. Not a collection of pages. An agent. This means: - A live chat that doesn't route to a human, it answers. It walks your buyer through their use case, configures a pricing plan, and qualifies whether they're a fit before escalating. - A configurator that's actually smart. Not "select plan A or plan B." An interface where the agent asks clarifying questions, makes recommendations, and builds a custom proposal in real time. - Intent capture that's conversational, not a form. A buyer shouldn't fill out 8 fields. They should have a conversation. The agent extracts intent from context. This isn't nice-to-have. This is demand gen infrastructure. Because every buyer who doesn't have a conversation with an AI now has a worse experience than your competitor who does. And the AI learns. Every conversation adds data. Your agent gets better at the close. Humans get the warm leads, not the cold suspects. **Pillar three: Autonomous campaign agents.** Your media planner doesn't allocate budget. An agent does. This means your marketing operations function shifts from "human makes campaign decisions and monitors them" to "human sets strategy and budget caps, agent optimizes within constraints." The agent: - Tests audience segments in real time and reallocates budget toward the ones converting - Iterates creative based on engagement data (not your opinion on what the ad should say) - Expands reach into adjacent audiences based on propensity modeling - Pauses underperforming channels and scales winners without a meeting A good autonomous campaign agent will outperform your best media planner 7 times out of 10. Not because it's smarter. Because it makes 10,000 micro-decisions a day while your planner makes 5. ## What Gets Disrupted (The Uncomfortable Truth) This stack doesn't just change tools. It changes headcount. Your email marketing team does email sends and list management. In agentic demand gen, these functions get automated or consolidated. A team of four becomes a team of one overseeing an agent. Or a team of one is reassigned. Your media planning team built campaigns and monitored performance. The agent does both. Your media buying headcount drops 50-70%. Your mid-funnel content ops team built nurture email copy, landing pages, and segment-specific messaging. Much of that automation moves to the agent. Some of the work survives (strategy, high-stakes campaigns), but the volume of creation shrinks. This is uncomfortable to say in an article. But it's true. The CMO who ignores this is ignoring the math. The CMO who manages this transition, retraining people, eliminating roles, and bringing in new ones, is the one who survives. ## What This Requires (New Roles, New Bets) You're not just buying tools. You're restructuring the function. **Role one: AI Content Ops Lead.** Your agent generates a lot of copy. Configurations. Answers. Not all of it is brand-safe. Not all of it is accurate. You need a person (or small team) who reviews, refines, and ensures quality. This person is a curator of agent output, not a generator. **Role two: Data and Attribution Engineer.** Your CAC attribution just got ten times harder. The buyer researched you in ChatGPT (you didn't see that). They talked to your site's AI agent (no traditional conversion event). They clicked a programmatic ad. They talked to your AI SDR. They opened an email. Three of those didn't exist five years ago. You need an engineer who can stitch this into a coherent attribution model and tell you, with confidence, what's actually driving pipeline. **Role three: Orchestration and Agent Infrastructure.** Your marketing stack used to be a stack. Email, CRM, analytics, ads. Now it's an organism where AI agents talk to each other, coordinate, and optimize together. You need someone (or a vendor, or a hybrid) who owns the plumbing. This is a new role. This isn't a RACI chart change. It's a function redesign. ## Your 3-Question Audit for Q2 Before you overhaul your entire marketing infrastructure, ask yourself three things: **Question one: What % of your traffic is now AI-assisted research?** Run a network analyzer. Pull your Semrush or SimilarWeb dashboard. How much of your traffic is direct referral that wasn't there two years ago? How much of your search traffic is compressed? Start measuring AI agent traffic specifically (through referrer analysis or direct instrumentation). If more than 30% of your buyer research is happening in an AI context where you're invisible, you're not optimizing for your real channel. You're optimizing for legacy infrastructure. **Question two: What does your brand look like when summarized by AI?** Actually run the prompts. Open ChatGPT, Claude, and Copilot. Ask them to compare you and your top three competitors. What does the AI say about you? Is your brand even mentioned in the analysis? If it is, is it accurate? Is it favorable? If an AI is summarizing your category and you're not in the answer, or the answer is stale, you have a content problem and a brand problem. This is your primary audience now. It's worth three hours of analysis. **Question three: If you cut paid media budget by 40% and invested the savings in agent infrastructure, what would break?** This is a thought experiment. But it forces you to think about where your real leverage is. If you cut 40% of ad spend and your pipeline drops 20%, you have a paid media problem. If you cut 40% of budget and pipeline only drops 5% because your agents are already converting demand, your paid media is a legacy cost center. This clarity changes everything about your 2026 plan. ## The Competitive Moat Here's what's not said enough: the CMO who still thinks of AI as a tool (a writing assistant, a chatbot, a reporting copilot) is already a quarter behind. AI is a tool if you're adding it to the old model. But the CMO who thinks of AI as the primary audience, the primary channel, and the primary coworker is building a moat. Because that CMO is saying: "What if I rebuilt demand gen to win in a world where my buyer researches through AI, buys through AI, and interacts with my company through AI?" And then they did it. That's hard. It requires new roles, new tools, new thinking, and the willingness to make some of your existing team's jobs obsolete. But it's also inevitable. The buyers are moving. The channel is moving. The only question is when you move with them. At Xivic, we spend our days helping brands do this transition. We've helped B2B and B2C companies rebuild their demand gen around agentic infrastructure. What works is not what you think. What works is radical simplification of the buyer experience, obsessive measurement of AI-assisted demand, and the courage to stop doing things that used to work. Your marketing automation platform will still exist in five years. But it will be a component of a much larger organism. Your CMO role, in five years, will look completely different. The sooner you start building for that future, the sooner you own it. --- # From ARR to ARR-per-Agent Source: https://www.xivic.com/xivic_insight_arr_per_agent.html > From ARR to ARR-per-Agent | Xivic Insights | insights and frameworks from Xivic. For 15 years, the SaaS growth playbook was ironclad: seat expansion drives NRR. Land a customer with 5 users. Expand to 15 within 18 months. Math checks out. Gross margin stays stable. ARR climbs. That playbook is breaking. In the last six months, we've watched leading SaaS companies report something that would have been scandalous a year ago: seat counts are *flattening* at their most strategic accounts. Not because of churn or competitive pressure, but because customers are consolidating seats. They're replacing human operators with AI agents. Intercom's new Fin product reports 65%+ resolution rates on customer support tickets, collapsing seat demand at top accounts. Klarna's CEO recently went public about replacing 700 full-time roles in two years, primarily with AI agents handling customer service, fraud detection, and collections. Salesforce is betting its entire future on Agentforce-a product category that explicitly positions agents as the unit of value, not users. Zendesk has quietly restructured pricing to meter per-conversation rather than per-agent seat. If your SaaS company is still measuring growth through seat counts, you're measuring the wrong thing. The metric that matters now is **ARR per Agent**-the revenue your product generates per autonomous workflow it runs for the customer. The companies that win the next cycle will be the ones that rebuild around this new unit of economics. The ones that don't will watch NRR collapse as customers consolidate seats or switch to agent-native competitors. ## The Uncomfortable Truth: Seats Are Not Your Growth Engine Anymore The traditional SaaS expansion playbook relied on a simple assumption: more users = more value. Each additional seat was a fixed economic unit. A $10/month per-seat model meant you could forecast expansion ARR with reasonable accuracy. Agentic AI broke that assumption. When a customer can deploy an AI agent to handle 80% of support tickets, or automate 60% of sales qualification calls, or run 500% more contract reviews with 20% of the human team, the seat-based growth curve flattens. It doesn't disappear, but it transforms. Top SaaS companies are already in this moment. They're not seeing this as a problem; they're seeing it as a restructuring opportunity. The winning move is not to fight it, it's to rebuild the business around the new unit of value. That unit is the agent. ## The Shift: From Users to Outcomes Old SaaS pricing model: $X per user per month. Expansion: more seats. New SaaS pricing model: $X per autonomous outcome per month. Expansion: more workflows, higher accuracy, broader autonomy. This isn't semantic. It's structural. When you price on seats, your customer's incentive is to minimize seat growth. They'll keep working with the same humans, just with better tools. When you price on outcomes, tickets resolved, proposals generated, customers qualified, the customer's incentive aligns with yours. They want the agent to handle more. They want you to expand the agent's scope. Intercom Fin is the clearest example. Rather than expanding Intercom's core per-seat model, Fin exists as a separate product tier where Intercom gets paid for agent-resolved conversations, not human login seats. The unit economics are different. The expansion mechanics are different. The customer's mental model is different. Salesforce Agentforce follows the same pattern. You're not buying agents as an add-on to your CRM seat count; you're building workflows and paying for outcomes. Even companies staying with per-agent pricing, like the new Zendesk model, are shifting the unit of billing from "how many people use this" to "how many agents does this run." It's subtle but decisive. ## The Metric That Actually Matters: ARR per Agent Here's how to think about your baseline today. Take your largest customer. What's their ARR? What's the number of agents (autonomous workflows, however you define it) your product is running for them daily? Divide one by the other. That's your ARR per agent. Now do that for your top 20 accounts. Average it. That's your benchmark. For most SaaS companies still operating on seat-based models, this number will be disturbingly low, $5K to $50K per agent depending on your vertical. That's because you're not optimizing for it yet. Your pricing, your product, your GTM are all still optimized around seats. Forward-thinking SaaS companies are already moving this needle. Outcome-based SaaS (support resolution, sales pipeline generation, recruitment funnel filling) are seeing ARR-per-agent metrics in the $100K-$500K range because the agent is doing the work that used to require multiple humans. But here's what matters more than the absolute number: **gross margin per agent**. Seats had predictable gross margins. You paid for your software, your infrastructure, your support, but the margin scale was linear. More seats meant proportionally more revenue with minimal incremental cost. Agents don't work that way. Agents burn compute. LLM calls cost money. Each additional workflow your agent runs increases your infrastructure costs. If you're not explicitly modeling gross margin per agent, you'll discover, too late, that you've scaled a negative-margin business. The companies that survive will be obsessive about this single metric: **How much gross margin (in dollars and %) are we making per agent we deploy for our customers?** That drives everything else. ## What Your Product Needs to Become If your product is still built around seats and workflows, you need a different architecture. An agent-native product does three things differently: **First, it becomes a platform for building agents, not a workflow tool.** Your product shouldn't be "use our UI to do X faster." It should be "deploy agents that autonomously do X and integrate with your stack." Slack is no longer a chat tool; it's an agent platform. Shopify is becoming an agent platform. Salesforce's entire bet is that CRM becomes agent infrastructure. The implication: your integrations layer becomes your moat, not your UI. Companies competing on design alone will lose to companies that can orchestrate across your customer's entire stack. **Second, human-in-the-loop becomes a feature, not a stopgap.** Early-stage agent companies treat human oversight as a safety layer, something you use until the agent is "good enough." Wrong frame. Forward-thinking product orgs treat human-in-the-loop as a permanent feature layer. Humans make the final decision on 5% of cases; the agent handles 95%. That's not a limitation, that's the product. **Third, your CS motion shifts from "helping customers use your product" to "helping customers expand their agent's scope."** Customer success becomes agent enablement. You're not asking "How can we get them to more seats?" You're asking "How can we get this agent to handle 20% more of their workflow?" ## The GTM Implications Your sales motion will compress even as ACVs expand. Traditional SaaS sales cycles: 4-6 months to land. Small initial ACV. 18-24 months to land major expansion. Agent-based SaaS: 2-3 months to land. Larger initial ACV (because you're replacing multiple seat subscriptions). 6-12 months to land major expansion (because expanding agent scope is faster than seat expansion). This means your sales team's job changes. You're not selling "let us help you work faster with these tools." You're selling "let us automate this entire function with an agent." That requires different hiring, different compensation (outcome-based rather than seat expansion), and different playbooks. Pricing strategy matters here, too. You have three levers: - **Per-agent pricing**: You charge for each autonomous agent deployed. Clean, scalable, customer-friendly. Risk: customers underdeploy if they're not sure about agent reliability. - **Outcome-based pricing**: You charge per resolved ticket, per generated lead, per closed deal. Perfectly aligned with value. Risk: complex to meter accurately; can create customer resentment if not transparent. - **Hybrid**: Seat licensing for the platform + outcome-based metering for agent use. Gives you optionality. Risk: confuses your GTM narrative. Most winning companies are landing with hybrid models then moving toward outcome-based as agents mature. ## Three Questions Every SaaS CEO Should Answer This Quarter If you're building or scaling a SaaS company, here are the three questions that will determine whether you're still viable in 18 months: - **What is our explicit agent-native product roadmap for the next 12 months?** Not "how do we integrate agents", that's table stakes. What is our core product's transformation into an agent platform? If you don't have one, your competitor will. - **What's our ARR-per-agent baseline today, and what's our target for 12 months from now?** If you don't know this number, you don't understand your new unit of economics. Calculate it this week. - **What's our NRR forecast if our top 20 customers replace 30% of their seats with agents in the next 18 months?** This is uncomfortable math, but it's the math that determines survival. If your answer is "our NRR collapses," then you need a different product strategy. If your answer is "our NRR actually improves because agents expand our total addressable scope," you're building the right thing. ## The Reckoning The SaaS companies treating agents as a feature, something nice to have once the core product is mature, are betting on a world that no longer exists. The SaaS companies treating agents as the core product unit, rebuilding pricing, GTM, product, and metrics around agent-based value, are building for the world as it is. This isn't a five-year transition. You're seeing it play out in real time. Intercom is already there. Salesforce is committing to it. Stripe is building toward it. Every category will follow. Your board should be asking about your ARR-per-agent metric next quarter. If you don't have an answer, you should start building one today. The SaaS winners of the next cycle won't be the ones with the most features. They'll be the ones with the most autonomous outcomes per customer. That's the game now. Build accordingly.