The Lion’s View built this checklist from the questions it applies in live engagements, evaluating AI and XR investments for private equity firms, family offices, and institutional investors before capital is committed. It covers four diligence dimensions: technical credibility, data and model dependencies, governance and regulatory posture, and commercial execution. Each section states the question a deal team should be asking, and the evidence that answers it — evidence that rarely appears in the pitch deck.
Technical Credibility
- Does the technology perform under the conditions the target market will demand, or only in a controlled demonstration?
- Can the company support its accuracy and performance claims with an independent test set, not the data it trained on?
- Does the architecture depend on a single proprietary model or vendor with no substitution path if that dependency breaks?
- Is the product’s current state consistent with what the roadmap and sales materials describe, or ahead of it?
- Has anyone outside the company’s own engineering team verified the technical claims?
Data and Model Dependencies
- Where does the training and inference data originate, and does the company hold defensible, documented rights to use it?
- What happens to product performance if a third-party data source changes its terms, pricing, or access?
- Does a documented lineage exist from raw input to model output, or is the pipeline undocumented and dependent on a small number of engineers?
- How exposed is the business to a single foundation-model provider’s pricing, availability, or policy changes?
- What is the company’s plan if the model version the product depends on is deprecated or materially changed?
Governance and Regulatory Posture
- Does the company have a documented AI governance framework, or does governance mean an informal engineering practice?
- Who owns model risk, drift monitoring, and incident response, and what is the escalation path when something goes wrong?
- Is the company positioned for the regulatory environment it will operate under in twelve to twenty-four months, not only the one in effect today?
- Can the company produce an audit trail for a specific model decision if a regulator, customer, or plaintiff asks for one?
- Does AI risk reporting reach the board, or does oversight stop at the engineering team?
Commercial Execution
- Does the pipeline reflect paying customers in production, or pilots that have not converted to revenue?
- What does the product actually cost to deliver at scale, and does gross margin hold once support and compute costs are fully loaded?
- How much of current revenue depends on a small number of design partners who received favorable terms to become reference customers?
- What does a customer’s measured outcome look like when verified independently of the vendor’s case study?
- Is the go-to-market model built for the buyer the investor is underwriting, or for a different market than the one in the thesis?
Before the First Conversation
In most engagements, two or three answers change the terms of the deal, not all twenty. The Lion’s View applies this framework directly in diligence sprints for PE and institutional clients; the questions above are the starting structure, not the full engagement. The reasoning behind each dimension — including what a data room typically hides and how the red flags differ by deal stage — is set out in full in the AI Due Diligence Framework for Private Equity.
Investors who want to talk through a specific target can review how The Lion’s View structures diligence engagements on the Investor Advisory page.
For how this framework applies in a specific sector, see Healthcare AI Diligence Beyond the Demo.
Get the board-ready PDF version of this checklist: