The AI exposure question every investment committee asks

Every investment committee now asks some version of the same question: if AI gets good at this, is the company still worth what we are paying? Around 60% of general partners have stepped up diligence on AI disruption risk, according to EY's Q1 2026 private equity pulse. The question is fair. The usual answer, a paragraph on "AI strategy" from management, is not evidence.

What does AI exposure mean in a deal?

It means the chance that AI makes part of the target's value cheaper to replicate or less necessary. That can happen to the product, if a customer can now get the same output from a general model. It can happen to the workflow, if the task the product supports gets automated away. And it can happen to the sales motion, if buyers start researching and choosing in ways the target's marketing does not reach.

These are different risks with different fixes, so they are tested separately.

How do you test the product?

Ask what the product does that a customer could not do with a capable general model and their own data. If the honest answer is a user interface on top of a workflow, exposure is high. If the product holds proprietary data, sits inside a regulated process or carries switching costs that come from integration, exposure is lower.

Look at what customers pay for. A product priced per seat for a task that AI can do in seconds will feel price pressure first.

How do you test the buyer's workflow?

Map the job the product is hired for and ask which steps AI can now do. If AI removes a step, the product either absorbs the saving and becomes more valuable, or loses the step and becomes less so. Interviews with customers and with the target's own sales team are the evidence here, because they say what buyers are asking for in renewals.

How do you test the go-to-market?

Check whether the sales team uses AI in ways that raise output, and whether the data underneath is clean enough for agents to read safely. Many companies have run pilots that never reached production. Score what is in production, not what was demonstrated.

Then check discovery. If buyers now ask AI assistants for vendor shortlists, does the target appear in the answers? That is a growing part of demand generation and a quick test of how current the marketing team is.

How should you score it?

Score exposure and readiness separately. A company can be highly exposed and well prepared, or barely exposed and unprepared. The first is manageable. The second is a risk that has not yet arrived. In our scorecard this sits in the Process and AI maturity area, scored 1 to 5 like the others, with its own data tests and conversations.

The AI Gap Analysis (five working days) goes deeper on how ready the revenue process is, and Deal DD covers exposure as part of the full scorecard.

Key takeaways

• Test AI exposure in three layers: product, buyer workflow, go-to-market.
• Defensibility comes from data, integration and distribution, not features.
• Score what is in production, not what was demonstrated.
• Score exposure and readiness separately.
• Tie the finding to price or conditions like any other risk.

Frequently asked questions

Is AI exposure only a risk for software companies?

No. Services businesses, marketplaces and data providers can all be affected. The three-layer test applies to any company that sells to businesses.

Can management's AI roadmap be trusted?

Treat it as a claim. Ask for what is live, who uses it daily and what measurable change it made.

How long does an AI exposure review take?

Inside Deal DD for GTM it is part of the fifteen working days. As a standalone AI Gap Analysis it takes five working days.