Services
AI due diligence
An independent view of whether claimed AI capability is real, defensible, and safe to buy.
Why it is its own review
AI claims now carry a meaningful share of valuation in software deals, and they are the hardest claims to check from a data room alone. A demo proves very little. The evidence sits in the models, the data rights, the telemetry, and the running costs.
Daniel has built and reviewed AI systems. The review starts from how the technology actually behaves, not from the pitch.
What gets examined
What the system actually is. Owned models, fine-tuned models, or a layer over a public model, and what that means for the valuation.
Where the data came from, who owns it, whether customers consented, and whether it can legally be used for training and improvement.
How the company knows its AI works. Test sets, benchmarks, and monitoring in production rather than claims from the demo.
Inference costs, routing, caching, and how the spend moves as usage grows. AI that works but loses money is a finding.
What happens when a model provider changes prices, terms, or behaviour. Dependence is normal. Unpriced dependence is not.
How prompts, outputs, and personal data are handled, and whether the AI surface creates exposure the rest of the product does not have.
Whether a competitor with the same public models could rebuild the product in a quarter, and where the durable value actually sits.
Prototype pipelines, unversioned prompts, missing evaluation, and the cost of making the AI maintainable after completion.
What you receive
Findings written for the deal team. Each finding is reported with the risk it carries, the commercial impact, and a recommendation. A briefing call follows the report, and the partners can question the findings directly.
Buying AI is one question. Governing it after the deal is another. For boards adopting AI inside the business, see AI governance and strategy.
Guides
Longer reading on how the work is done.
Before you invest, sign, or build, talk it through.
Book a confidential call. Bring the decision. Daniel will say how he can help, or tell you plainly if he cannot.
Book a call