What makes AI stick in real estate.

This is the first RE·AI newsletter. Each issue sets out what I am seeing on live engagements: what works, what does not, and why. It is short by design.

RE·AI is a specialist AI consultancy that works only in real estate. I founded it after a career in investment, asset management and advisory, most recently as a Vice President at Brookfield Asset Management. The reasoning behind it is straightforward: a system that drafts an IC memo or a comps schedule has to be designed by someone who knows what a good one looks like.

Three observations

01

Workflow fit

Around 90% of real estate companies are piloting AI. According to JLL, only 5% have achieved all of their AI goals. In my experience the gap is rarely the model. It is where the tool sits.

A platform that asks an analyst to leave Excel, re-key a deal and come back again is abandoned within weeks. The systems that last work inside the tools already in use: the house model, the IC template, Outlook. The only change the team notices is that the first draft arrives sooner.

02

Defining the standard

Before anything is automated, the standard has to be written down. Most firms carry it in the heads of two or three senior people: which comps are excluded, how ERV is evidenced, what an IC memo must cover before it goes to committee.

Once it is written down, the system can be checked against it line by line, and every number traced back to its source. Without it, nobody trusts the output, and an output nobody trusts is redone by hand. Every output remains a draft until a named member of the team approves it.

03

Proving it on a live deal

I recommend starting with one workflow, run on a real transaction, rather than a firm-wide pilot programme. A live deal shows the hours saved, the errors caught and where the process needs adjusting within a few weeks. Scale follows the evidence.

MIT's 2025 research points the same way. It found that 95% of enterprise AI pilots produced no measurable return, and that tools built with specialist partners succeeded c.67% of the time, with internal builds succeeding about a third as often.

From the build

A pan-European logistics investment manager. The system reads each investment memo, fills in the screening chart and the cashflow model, and shows where every number came from, drawing on a store of 7,676 comparable transactions. The recoverable time is c.380 analyst hours a year. The client's name is withheld.

c.380

Analyst hours a year