95% of AI projects in real estate are failing.
Not because AI doesn’t work. Because it’s being used for the wrong things.
In 30 years working inside real estate portfolios, here’s what I’m not seeing talked about enough.
Boards are pressuring real estate companies to pursue AI. But most are skipping the foundation.
AI is brilliant at probabilistic work: summarizing documents, extracting meaning, scoring risk. It is not a replacement for clean, structured data.
A client recently acquired a significant portfolio. 10,000 documents needing due diligence review. The traditional process would have taken months – left the team drowning in spreadsheets, chasing version control, second-guessing completeness.
We did it differently.
AI extracted metadata and summarized key findings for every document. It generated a risk score for each one. Humans reviewed and validated those scores – which is exactly where human judgment belongs. A relational database tracked everything, flagged outliers, and kept the team on one source of truth.
The result:
↳ Review time cut by over 90%
↳ Accurate, complete risk scores for every document
↳ One source of truth the whole team could rally around
↳ A production-ready dataset from day one of ownership
That last one rarely gets mentioned. When the deal closed, they didn’t spend the next six months trying to understand what they’d bought. They already knew.
That’s what happens when you let AI do what AI is good at, and databases do what databases are good at.
If you’re being asked to “do something with AI” and it isn’t working, this is probably why.
