Blackstone deal committee reads AI memo on most deals

Blackstone's Monday investment committee now receives two memos on nearly every deal. One is from the deal team. The other is from a model trained on more than 14 years of the firm's internal deal memos and debates, according to Commercial Observer.

The model has no vote but has a voice in the room. Bloomberg last month called it the firm's artificial intelligence chief investment officer, Commercial Observer reported.

Dealpath's 2026 survey of 103 investment and technology professionals at institutional commercial real estate firms finds adoption ahead of usefulness. Ninety-seven percent said AI is integrated into their investment process. Fifty-one percent said it saves time after the cost of checking its work is counted.

For an allocator hearing an underwriting-speed pitch, the useful question is whether the time savings survive the checking and whether the model was trained on the firm's own deal history.

Fifty-five percent highly trust AI to summarize a diligence document. Thirty-five percent trust it to score and rank deals, the lowest share of any task the survey asked about.

Commercial Observer says the gap comes from what the model can see, not which model it is. Every firm can license the same software and gets the same upgrade on the same day. A summary needs one document that holds everything. A deal score needs the buy box, the deals a firm pursued and passed on and why, the comps it trusted last time and what else is trading now. A general-purpose model has seen none of that, so its score is a guess at someone else's strategy.

Size shows up in the answers. Professionals at firms managing $40 billion or more rate their data the most AI-ready of any group and extend AI the most trust on judgment work, according to Commercial Observer. The pattern is data investment first, trust after.

The survey shows both high readiness and a limiting data problem. Eighty-three percent of respondents rate their firm's data infrastructure mostly or fully ready for AI, and 90% say data quality or fragmentation has limited AI's impact at their firm. Both can hold at once because 94% said they check the model's output most of the time or always. That verification tax is where the promised time savings go.

Apollo's Torsten Slok has spent the year warning that outside technology there are still no signs of AI lifting profit margins and that data governance could delay the gains for years. The survey shows commercial real estate adoption near-universal while timing bears him out.

For managers raising capital, the verification tax is where they will have to make their case.