DD vendors now use AI for the standard part of legal and tech reviews, and their fees haven’t moved. A partner at an investment firm put it to me that way a few weeks ago. He assumes most vendors use AI extensively by now, and he doesn’t see any price cuts.

Shortly before that, I had asked a DD lawyer how they handle the standard part of a legal review. Their proprietary AI takes care of that scope, they told me. My guess is that it’s a fairly advanced Claude skill, which is also what I would build in their place.

What the standard part actually covers

A large part of every report was the same checklist applied to a new data room. For legal that means clauses like change of control and indexation, and for tech it’s things like licenses and key-person risk. In some cases, that type of standard work made up most of the hours.

The work was never difficult, it just took a long time. Someone had to open every contract and record the same handful of fields, and the hours added up because the data room was big.

AI now produces that part in minutes.

Why the price should move

Once the standard work costs almost nothing to make, there is little left to justify its price. That goes for every service that sells an assessment, from audits to compliance reviews.

The fee for that scope used to pay for time. If the time is gone and the fee stays, you are paying for something the vendor no longer spends.

Splitting standard scope from judgement scope

For a PE fund this raises a practical question. Can you split the standard analyses from the custom ones and pay for them differently? You could run the standard sweep yourself and give the advisor a narrower brief that focuses on the judgement calls.

In practice that means writing the brief in two parts. The first part lists the items you expect to be mechanical, such as the clause checks, the licence inventory and the contract register. The second part lists the questions where you want someone’s opinion on the record, such as whether a customer concentration is a real risk at this price.

You can also ask the vendor to price the two parts separately, even if they end up doing both. That alone tells you how they see the split.

There is a limit to how far you can push this. If you run the standard sweep yourself, you own the result of it, and the advisor will scope their opinion around whatever you handed them. That trade is worth making on a deal you know well and worth skipping on one you do not.

The pushback, and where it lands

Advisors won’t like that. I can already hear the response: “I don’t want to argue with the bullshit your Claude skill puts out.” They have a point, since they carry the liability and reviewing someone else’s output takes time as well.

Reviewing work you did not produce is genuinely slower than producing it. And the signature at the bottom of the report carries a cost that has nothing to do with how long the analysis took.

Still, the pricing argument stands. If the vendor uses AI for the standard part, the fee for that part should go down. That applies to the scans I sell too.

A test before your next assessment

Next time you commission an assessment, run the standard items through Claude yourself before the report lands. Whatever the advisor found that you missed is what you’re actually paying for.

It is a cheap experiment and it gives you a number to negotiate with.

Originally posted on LinkedIn