The deck says proprietary AI. In three to five days you know what you are actually buying.

Whether the target’s AI is real, what it costs to keep running, and what walks out of the building when three people leave.

For private equity and venture funds, family offices and M&A boutiques. It runs on the deal’s timeline, it ends in a verdict, and when the deal closes or dies, so does the work.

It can also run as one workstream inside a wider diligence led by your lawyers or your financial advisers, on their timeline and in their data room.

For targets in Romania and the region there is a practical advantage, and I’d rather state it plainly than hint at it: I read the data room in Romanian, I know which local vendors deliver and which do not, and I have worked inside retail, banking, manufacturing and energy companies, on the payroll, and that is a different vantage point from advising them.

Three levels, nested

Level	Scope	Typical use
1 – Red-flag screen	Three to five days. An eight-signal gate returning GO, GO WITH FOCUS, or STOP.	Before the letter of intent, when access is limited and a full diligence is not yet justified.
2 – Standard	Full workstream coverage with scoring and evidence.	Confirmatory diligence, where AI is material to the thesis.
3 – Deep	Extended evidence, management testing, cohort assessment of the target’s team.	Large tickets, or deals where the AI is the investment thesis itself.

The Level 1 fee is credited in full against Level 2. If the screen says stop, you have spent a few days to avoid a few weeks; that is a good trade even in the cases where the deal would have been fine.

What gets examined

• Whether the AI is real or a wrapper with a story: tested, not asked.
• Who holds the rights to the data the models were trained on, and what happens to those rights at closing.
• Key-person risk. How many of the people who understand the models are on the org chart, how many are contractors, and what the retention terms look like.
• The whole AI estate, including the parts that predate the generative wave: classical machine learning, deep learning, computer vision, video, audio, forecasting. Their economics have nothing to do with tokens.
• The embedded AI inside the target’s SaaS stack: what is switched on, whether it can be switched off, who controls updates, what it costs, and where the data goes.
• Whether the unit economics hold at volume, or only in the deck.

What you get

One page with the verdict, the short list of what to verify if you continue, and the evidence behind each finding. The full scorecard sits underneath and you get all of it; I would be surprised if anyone on the investment committee reads past page one, and the page is written on that assumption.

The management team, on the same scale

Where it matters, the maturity diagnostic runs on the target’s own executive team as evidence inside the diligence, graded on the same twelve-grade scale as everything else. It tends to be the part the deal team remembers, because a management team that cannot agree on where it stands today is unlikely to deliver the AI plan the valuation assumes.