The AI literacy due diligence checklist your deal team is missing

Every pitch deck now includes a slide about AI. Most of those slides mean nothing. Here is how to tell the difference between a company that has real AI capability and one that has a ChatGPT subscription and good marketing.

1. Does the team build with AI or just use AI?

Ask the team to describe their AI workflow. Companies that use AI will talk about tools: "We use Copilot for code suggestions" or "Our marketing team uses ChatGPT for first drafts."

Companies that build with AI will talk about systems: "We built an internal reporting tool using Claude Code that pulls from our database and generates the compliance report our auditors require."

The difference matters. Tool users depend on vendors. Tool builders own their capability. When a vendor changes pricing, terms, or features, users are exposed. Builders adapt.

2. Is there a governance document?

Ask to see the AI policy. Not a slide about AI strategy. A written document that says: what AI tools are approved, who approved them, what data enters them, and who reviews AI outputs before they are acted on.

A CLAUDE.md file in the codebase counts. A written AI use policy counts. An internal wiki page with clear rules counts.

What does not count: "Our CTO handles that." If it is not written down, it does not exist as governance. It exists as one person's opinion, and it leaves with them.

3. Can non-technical leaders demo what they have built?

Not present a slide deck about what AI could do. Actually show a tool. Open a browser, navigate to an internal application, and demonstrate: "This is what it does. I helped build it. Here is how it works."

If the CEO cannot do this, the company has AI tools. It does not have AI capability. The distinction is the difference between a company that hired a contractor to install solar panels and a company whose team understands their energy infrastructure.

4. Is their AI stack owned or rented?

Ask where the AI code lives. If the answer is "in our GitHub repository, deployed on our infrastructure," that is ownership. If the answer is "we use [Vendor] and they handle everything," that is a rental agreement.

Rented AI stacks create three risks: vendor lock-in, data portability limitations, and pricing vulnerability. A company whose core workflow depends on a SaaS AI tool is one pricing change away from a margin problem.

5. Do they have data governance before AI deployment?

Most companies deploy AI tools before understanding their own data. The AI then operates on data that is inconsistent, incomplete, duplicated, or incorrectly categorized.

Ask: "What data quality process runs before data enters your AI tools?" If the answer is silence, the AI's outputs are built on an unreliable foundation. That is not a technology risk. It is an accuracy risk that compounds over time.

6. What happens when the model changes?

AI models update. Sometimes they improve. Sometimes they break workflows that depended on specific behaviors. Ask: "If your primary AI provider releases a new model version next month, what is your process for testing, validating, and migrating?"

Companies with mature AI practices have version pinning, testing protocols, and fallback plans. Companies without them are running production workflows on whatever the vendor ships, whenever the vendor ships it.

The valuation argument

Companies that score well across these six questions are not just better AI users. They are structurally more defensible. Their capability is internal, governed, and transferable. Their risk exposure is documented and managed. Their dependency on any single vendor is limited.

That is a valuation argument, not a technology argument.

Talk to us about portfolio AI readiness — we work directly with investors and portfolio companies on AI readiness assessment and team training.