The written answer.

A new question has been showing up in due-diligence questionnaires. When AIMA last surveyed managers and their investors, 29% of allocators already had specific AI questions in their operational diligence and another 29% expected to add them within the year; the follow-up study is in the field now. Mercer’s 2026 numbers suggest what it will find — 55% of managers with AI in at least one investment process, 91% planning more — and Mercer’s ODD group has started calling this an inflection point, noting diligence findings tied to technology deficiencies rose 5% in a year. So: nearly everyone is using it, and a growing share of institutional investors want the details in writing.

The question looks technical. It is not. Which tools, under what terms, touching which data, watched by whom — that is the same grammar diligence has always used for brokers, administrators, and valuation policy. What the allocator actually wants to know is whether the shop runs deliberately or by accretion. AI just happens to be the newest thing in the building, which makes it the easiest place to check.

At most funds the honest answer is currently unwritten. People use what they use; somewhere there is a policy memo; the two are different documents, and everyone involved knows it. That gap survives a pitch meeting. It does less well in an ODD review, where the operational answer is the thing being examined and “we are still formalizing our approach” tends to read as a description of the whole shop.

Here is what makes the question worth taking seriously rather than handing to boilerplate: a good answer is short. An approved set of tools, chosen on purpose. Provider terms settled where they matter — retention and training — in writing. A plain account of which documents may enter which systems. A name attached to all of it. One page. Deliberate practice is describable; improvised practice resists description, which is why the improvising firm produces either nothing or eleven pages of fog.

Firms that sit down to write the page tend to discover the writing improves the thing described. The inventory turns up personal subscriptions nobody had mentioned. Settling terms forces the retention conversation that kept getting deferred. Sorting documents into systems is, for many small firms, the first real data-classification exercise anyone has done. The questionnaire is the occasion. The clarity is the point.

One caution. The failure mode here is the policy that outruns the practice — controls that exist on paper, produced for the questionnaire, contradicted by the browser history. That is worse than no answer, because it turns an operational gap into a candor problem. Write the page downstream of what actually happens, even if that means changing what happens first.

The question will become standard the way cybersecurity questions became standard, and then it will stop differentiating anyone and start being a floor. The managers who stay comfortable with it will be the ones whose answer was written by their operations rather than for their marketing — the ones who could produce the same page next year, changed only where the practice changed.