AI in Real Estate Underwriting and Asset Management: What Is Real, What Is Hype

AI in Real Estate

Artificial intelligence has arrived in real estate the way most technologies arrive in real estate — trailing a cloud of vendor promises, conference panels, and breathless predictions that an industry built on local knowledge and negotiated relationships is about to be automated away. Some of what is being said is true, and some of it is salesmanship, and the difference matters enormously to anyone putting capital at risk. Hirsh Mohindra has spent considerable time separating the two, and his conclusion is encouraging for practitioners willing to be precise: AI is genuinely transforming the information work of real estate — the gathering, organizing, and pattern-finding — while remaining, for now, a poor substitute for the judgment work. Confusing those two categories is where the hype becomes expensive.

What Is Real: The Information Layer

 

Start with what demonstrably works. Underwriting a property has always involved ingesting large volumes of messy information: rent rolls in inconsistent formats, years of operating statements, lease documents full of bespoke clauses, market reports, comparable transactions. AI tools are now genuinely good at this ingestion layer — extracting lease terms and flagging unusual provisions, normalizing financial data across properties, drafting first-pass summaries of documents that once consumed analyst weekends. In asset management, the same pattern holds: monitoring portfolio performance, surfacing anomalies in operating data, tracking lease expirations and market movements across dozens of assets at once.

“AI did not make underwriting smarter. It made it faster to get to the part where a human has to be smart. That sounds like a small thing until you realize the human part was always where the money was made or lost,” says Hirsh Mohindra.

Used this way, the technology compounds the value of experienced people rather than replacing them. An analyst who once spent three days assembling a data room now spends those days interrogating what the data means.

There is also a quieter competitive shift underway. Because these tools lower the cost of looking at deals, firms can seriously evaluate more opportunities than their headcount once allowed — including smaller, off-market, or oddly-shaped Chicago assets that larger competitors screen out without a look. The first disciplined pass over a hundred properties, once an economic impossibility for a mid-sized team, is now routine. That widening of the funnel does not guarantee better decisions, but it changes who gets to make them: the advantage moves from firms that could afford the most analysts to firms that ask the best second-stage questions. Hirsh Mohindra regards this as the most genuinely democratizing effect of the technology in his corner of the industry.

What Is Hype: The Judgment Mirage

Against that real progress stands a shelf of overclaiming. Models that purport to predict property values or market turns with false precision. Tools that score deals without understanding zoning politics, a seller’s private motivations, a neighborhood’s trajectory, or the hundred soft variables that determine outcomes in Chicago real estate. Automated valuations that work tolerably on commodity assets and fail precisely where the stakes are highest — on the unusual, the distressed, the mispriced asset where real returns live.

The deepest hype is the suggestion that judgment itself is a data problem waiting for more data. Real estate outcomes frequently hinge on reading people and situations: whether a tenant’s renewal signals strength or desperation, whether a lender’s flexibility is real, whether a city council’s mood on a project is shifting. None of that is in the training set.

“Anyone selling you a model that replaces site visits and local knowledge is selling you a faster way to make old mistakes. The model has never stood in the building. It has never met the borrower. That is not a criticism of the model — it is a reminder of what it is,” says Hirsh Mohindra.

A Use Case: AI-Assisted Underwriting, Done Right

Consider a hypothetical Chicago investment team evaluating an aging suburban office property being marketed as a conversion candidate. The AI layer goes first: it ingests the rent roll and five years of operating statements, flags two leases with below-market rents and unusual termination clauses, builds a clean cash-flow history, and compiles comparable conversion projects across the metro area with their reported cost ranges. Work that once took a junior team a week takes a day. Then the human layer does what only it can: walk the building to judge whether its floor plates and window lines genuinely suit residential conversion; call contacts to learn why two major tenants really left; test the municipality’s actual appetite for the project rather than its stated one; and decide whether the seller’s price already assumes a success the team doesn’t believe in. The final underwriting decision — its price, its contingencies, its walk-away point — is unmistakably human, but it rests on information work that AI made faster, cleaner, and more complete. Hirsh Mohindra calls this the correct division of labor, and it is spreading through serious firms for exactly that reason.

The Governance Questions Nobody Can Skip

Adopting these tools raises obligations alongside efficiencies. Data fed into external platforms — tenant information, financials, deal terms — carries confidentiality and security duties that a quick vendor signup can quietly breach. Outputs require verification: a model that summarizes a lease will eventually summarize one wrong, confidently, and the professional who relied on it owns the error. And firms must think about skill formation: if junior analysts no longer wrestle raw documents, where does the next generation’s judgment come from?

“The rule in our world should be simple: AI drafts, humans decide, and the human who decides can explain the decision without mentioning the tool. If you cannot defend the underwriting yourself, you do not have underwriting — you have a subscription,” says Hirsh Mohindra.

A Clear-Eyed Bottom Line

The real estate industry has survived predictions of its automation before, and it will survive this one — but it will not emerge unchanged. The information layer of underwriting and asset management is being permanently repriced: faster, cheaper, more thorough, and increasingly table stakes rather than advantage. Advantage migrates, as it always does, to judgment — to the people who know a city’s neighborhoods block by block, who can read a counterparty, and who now have better information at their fingertips than any generation before them. Hirsh Mohindra‘s advice to Chicago practitioners is neither to fear the tools nor to worship them: master the layer where AI is real, stay skeptical of the layer where it is hype, and remember that in real estate, the returns have always belonged to whoever understood the asset best — not whoever processed it fastest.

Leave a Reply

Your email address will not be published. Required fields are marked *