The AI Exit Story: How to Underwrite AI Value at Exit

By the time your portfolio company reaches exit, AI has likely driven material value creation—or destroyed significant upside. The question is whether you can prove it. Buyers today are sophisticated enough to ask: what does this portfolio company's AI actually cost to run, and what is the true economic contribution? This article walks through how to structure the AI exit story in your diligence book, craft the management presentation narrative, and use work-item-level cost attribution as proof of operational excellence.

The AI Exit Narrative Buyers Are Asking For

Ten years ago, a PE exit narrative centered on headcount reduction, margin expansion through cost discipline, and revenue synergies. Today, buyers want the AI story. Did the portfolio company deploy AI to eliminate work steps, automate customer interactions, or unlock new margin? Is that AI running on a sustainable cost model, or is the company burning cash on inference? Can you quantify the AI-driven margin contribution per transaction, per claim, per customer interaction?

According to McKinsey's State of AI 2025, 88% of organizations use AI in at least one function, but only 39% see measurable EBIT impact. Your job as an operating partner is to make sure your company sits in that 39%—and that the buy-side due diligence team can see it clearly. The best AI exit stories show concrete work metrics: "this claims team was processing 120 claims per FTE per month; with our AI-assisted workflow, they're now processing 180 claims per FTE with the same headcount, and AI cost us $18 per adjudicated claim."

That is not a cost center. That is a value driver.

Building the AI Section of the Diligence Book

Your diligence book should have a dedicated AI economics section, separate from IT infrastructure. Structure it around three questions buyers will ask: (1) What is the portfolio company's true AI spend? (2) What is the cost per work outcome? (3) Is there lock-in risk, or can the company unwind the AI and maintain the operational improvement?

Start with the cost transparency piece. Most portfolio companies live in stage 1 or 2 of the 5-Stage AI Cost Maturity Curve—they know they're spending on AI, but they can't break it down by application. By exit, you should be at stage 3 or 4: able to show AI spend allocated to specific business units and, better yet, tied to specific work items.

The diligence book section should include:

• Itemized AI spend breakdown: APIs (OpenAI, Anthropic, Google), third-party service providers (Zapier, Make, n8n), internal model inference, vector database and storage, observability (Langfuse, Helicone), human review and quality assurance, training and validation data.
• Work metric trend lines: How have claims processed per FTE, customer interactions resolved per agent, loan applications approved per officer changed since AI deployment?
• Cost-per-outcome KPI: This is the magic number. "AI-assisted claims adjudication costs us $18 per claim, and the prior manual process cost us $32. The delta is $14 per claim, on a 280,000-claim-per-year volume, equals $3.92M of annual value creation."
• Vendor and contract summary: Which third parties run critical workflows? What are the contract terms, lock-in clauses, and termination costs? Is the company dependent on a single vendor, or is the AI setup portable?

The AI Cost Iceberg—visible API spend over hidden inference, retries, observability, and human review—is essential context here. Buyers will ask why the OpenAI bill is $12K per month but the all-in AI cost is $87K per month. You need to walk them through the hidden costs and explain why the $87K is actually the right number to use when modeling sustainability.

The Management Presentation: Talking Points for the AI Story

When your CEO and CFO sit across from a sponsor bank or strategic buyer, they will face detailed questions about AI. Here are the talking points your management team should be ready to deliver:

Point 1: AI as an operational lever, not a cost center. "We didn't adopt AI to reduce headcount. We adopted it to redeploy headcount from manual to higher-value work. Our claims adjudication team grew by 12% in the past 18 months, but processed volume grew by 38%. AI-assisted workflow allowed us to do more with modest headcount expansion."

Point 2: Cost-per-outcome is sticky and defensible. "Our cost per adjudicated claim is $18, and it's stable across the portfolio. Unlike raw headcount reductions, which are easy to unwind, the workflow improvement is embedded in standard operating procedure. A buyer can continue running this playbook and realize the same margin expansion we did."

Point 3: Vendor risk is managed. If you've built AI on third-party APIs only (OpenAI, Anthropic, Google), you can say: "We use commodity APIs, not proprietary models. If we wanted to migrate to a competitor, we could, in weeks. Our workflows aren't locked into a single vendor." If you've built custom models or integrated deeply with a single vendor, be transparent: "We're integrated with [vendor], which gives us the cost advantage. A buyer would inherit that relationship—or could migrate at a known cost."

Point 4: The margin story compounds. "Today, AI is driving 2.8 basis points of gross margin improvement per transaction. If a buyer can apply the same playbook across [related business unit or geography], the opportunity is materially larger. The AI setup scales."

Cost Attribution as Proof of Excellence

Here's where work-item-level cost attribution becomes your secret weapon. Buyers (and their advisors) will be deeply skeptical of any claim that "AI costs $18 per claim" without seeing the ledger. If you can show them a Runrate-grade cost attribution system that ties every dollar spent on inference, retries, and human review back to a specific claim, you've just moved from story to proof.

The diligence narrative becomes: "Our finance team runs daily cost attribution on 100% of our AI-assisted workflows. Here's the dashboard: claim ID 4821847 cost $17.34 in AI infrastructure. Here's the composition: $3.82 in GPT-4 inference, $2.14 in vector DB retrieval, $1.99 in human review, $0.51 in error retry. We have 18 months of daily data. We're not guessing. This is auditable."

This is also where you neutralize the buyer's inevitable concern: "What if AI stops working tomorrow?" Your response: "We know exactly what happens. Every percentage point of AI accuracy that drops adds $0.23 per claim in manual review cost. We've stress-tested it. A buyer can model their own tolerance."

Planning the AI Exit Timeline

The best AI exit stories are built over 12–18 months of operating partner engagement, not three months of diligence scrambling. Here's the timeline:

Months 1–3: Audit AI spend and cost attribution. Get to stage 3 maturity minimum (AI spend allocated by business unit). If you're starting from shadow AI and ad-hoc spending, this takes investment.

Months 4–8: Establish cost-per-outcome KPIs and build the trend-line story. Buyers want to see 6+ months of data showing stability or improvement, not a one-off snapshot.

Months 9–12: Begin drafting the AI section of the diligence book. Have your CFO and COO practice the talking points. Stress-test the numbers with your transaction advisor—if they're skeptical, fix them now.

Months 13–18: Refine the narrative based on latest data. By exit, you want the AI cost attribution data so clean that the buyer's IT advisor can review it in a single day and say, "Yep, this is real."

The Sustainability Test: Can a Buyer Maintain This?

The ultimate buyer question: is this AI-driven margin improvement something the buyer can maintain, or does it disappear with your operating partner? Frame it this way:

• If the AI relies on custom models or proprietary training data: The buyer inherits a risk. Be transparent about the investment required to maintain the advantage. Position it as value-creation optionality for the buyer, not a liability.
• If the AI relies on commodity APIs and standard workflows: This is gold. You can say with confidence: "A buyer can run this exact playbook. The margin profile is stable and independent of our involvement."
• If the AI requires specialist resources (a data engineer, an ML ops person): Estimate the all-in cost of that headcount and embed it in the cost-per-outcome KPI. A buyer needs to know what they're inheriting.

The strongest exit narratives show that AI-driven margin improvements are operational and defensible—not dependent on individuals or secret sauce.

From Diligence to Valuation

Here's the honest truth: buyers may not pay a premium for AI margin improvement alone. But they will absolutely discount you if you can't explain it. A buyer who finds material hidden AI spend—or discovers the AI driving the claimed margin improvement is unsustainable—will use that discovery to renegotiate purchase price.

Your job is to make the AI story so clear and quantifiable that it removes a source of uncertainty. The buyer's advisors will stress-test the AI cost attribution. The more bulletproof your diligence book, the faster that process moves, and the less room for discount negotiation.

By the time your company reaches exit, AI should be woven into your operational fabric. The exit story is simply transparency about the work that happened over the past 24 months. For more on how to build this operational discipline during your holding period, see the PE Operating Partner AI Playbook. When you're ready to see how work-item-level cost attribution accelerates this process, request a demo.