Can AI Deliver Real-Time Risk Warnings For CRE Portfolios?

If Walgreens shuttered a swath of stores tomorrow, a new tool fueled by generative artificial intelligence is claiming the ability to outline a swift course of action for landlords stuck holding the keys — the latest in a wave of new AI-fueled commercial real estate products rushing to market as a once-tech-shy industry embraces a more data-driven future. 

Earlier this month, Moody’s launched Early Warning System, an AI-powered tool aimed at alerting CRE clients to real-time risks to their portfolios based on breaking news such as large companies downsizing physical footprints or filing for bankruptcy. The tool claims to offer immediate information on personal exposure to events, key metrics on affected properties, and Moody’s credit opinion of the company.

Early Warning System aims to help CRE companies more effectively and quickly determine the impact of such events and streamline the process of determining the path forward, said Joe McBride, Moody’s senior director of CRE product management.

“If we can bubble that up for that portfolio manager and show it to them in real time, then the next steps could happen a lot sooner than they might happen otherwise,” McBride said. 

Moody’s product, one of the first real-time, AI-based CRE tools to enter the market, is part of an AI arms race in which industry players are attempting to harness the soaring potential of the technology. 

JLL became the first major CRE brokerage to launch its own AI-powered large language product in August 2023, dubbed JLL GPT. Later that month, CBRE announced that it was using AI to help maintain 1B SF of its portfolio. 

Moody's says its tool takes the technology further. It could have the timing right.

A June Colliers survey found that 71% of top leaders plan to implement their first generative AI solution within two years, estimating that activities accounting for as many as 30% of hours currently worked across the U.S. economy could be automated.

But some are still in the dark about how it works.

McBride walked Bisnow through a product demonstration with the same sample portfolio that includes Walgreens locations to detail how a client could utilize it in practice.

Even if Walgreens closing locations didn’t impact a particular client’s properties in the immediate term, the product is capable of identifying stores potentially at risk in the future, McBride said. It can give an idea of properties that could be targeted in future rounds by identifying locations where rent is higher than the market rate, making them prime candidates for the chopping block.

The tool could also sift through which leases are soon to expire and which are located in less desirable, aging buildings, giving owners a heads-up that they could be in trouble.

Users can utilize the large language models built into the product to work through different scenarios, such as asking the LLMs how net operating income and debt service coverage are impacted if a portfolio owner has to re-lease properties Walgreens leaves at a lower rent, McBride said. 

Moody’s designed the product to help clients get ahead of potential property issues or distress, McBride said. A client’s next steps following a piece of news relevant to its portfolio might include rerunning a risk screening analysis, adjusting an allowance on a loan or calling a borrower and finding out the details behind a story, he said.

To develop the comprehensive database that fuels the product, Moody’s acquired several companies in the data space over the past several years, built out underwriting and origination products, and expanded its underlying CRE data set, McBride said. The company also built a tool that aggregates news from across the world and puts news items into a data feed, which Moody’s integrated into the platform to tailor alerts to specific clients, he said. 

“The real-time nature, I think, gets you ahead of that renegotiation or refi or whatever stress or potential friction that is going to occur or may occur,” McBride said. 

Moody’s stress-tested the system through a host of scenarios to ensure the AI’s accuracy, checking that manual calculations match what comes out of the tool, McBride said. The tool’s refinement will be an ongoing process as users ask it questions the team didn’t initially think of, he said. 

“​​The most important thing is that anything to do with math, anything to do with asking a question about the portfolio information, we need 100% accuracy, and that's what we have so far,” McBride said. 

McBride said Moody’s product differentiates itself from competitors in the generative AI space by folding in its research on corporate credit, the news aggregation tool and a sizable amount of CRE data. Clients are also already uploading rent rolls, operating statements and loans into Moody’s systems, which gives the company a “bit of a head start,” McBride said.