Artificial intelligence (AI) is expected to enable a dramatic leap in the commercial real estate industry’s analytical and predictive capabilities, but a significant amount of fastidious work to wrangle data and train generative models is foreseen before those new efficiencies and insights can be achieved.
Early adopters and leading enthusiasts tally a long list of possibilities for portfolio management, risk management, valuation and assessing investment performance. For now, though, they’re mostly slogging through the monumental task of gathering and integrating the underlying intelligence.
“We’re spending our time cleaning client data so we can feed it into our models to provide the analytics back to them,” Charles Fisher, JLL’s director of global real estate risk analytics, reported earlier this year during a webinar sponsored by the Open Standards Consortium for Real Estate (OSCRE). “To get to scale, it gets more complicated. Definitely, there are lots of levels of maturity on this curve.”
Eventually, he anticipates AI will be enlisted in that task, once models can identify and remove rogue data. In the interim, the industry is accumulating and refining resources, and employing a range of application programming interfaces (APIs) to integrate data from across its multidisciplinary landscape to spot trends, forge connections and inform decision-making. Firms are also implementing AI where it’s workable.
“We’re already using generative AI to update some of our repetitive tasks like summarizing documents, data collection, aggregation, report generation or even some of our more administrative tasks like taking notes during meetings, summarizing the outcomes, takeaways and action items,” Amanda Carrillo, director of analytics insights and intelligence for CBRE Investment Management in the Americas, told the webinar audience.
Kevin Shtofman, global head of innovation with the real estate data and analytics consulting firm, Cherre, highlighted some of the machine learning functions — which he characterized as “the step before you get to AI” — that have become fairly commonplace. For example, it can serve as a swift and meticulous proofreader, cross-referencing the many contributing details in companies’ reports with the source documents.
“It’s automatically running some validation rules around lease length, correct lease dates, correct square footage, correct number of units, matching and netting trial balances,” Shtofman explained. “It’s just applying a lot of basic rules to prevent a flawed dataset from being presented.”
He also described some early inroads in using AI to predict the risk of tenants defaulting. Risk profiles are derived from history of rent payment timeliness and occurrences or patterns of falling behind and requesting restructuring of payment terms, which are analyzed across all leases a tenant holds within a landlord’s portfolio.
“Understanding AR (accounts receivable) risk has been the first use-case where we’ve found a lot of success in AI,” Shtofman said.
Looking to the future, webinar participants noted both generative AI’s interface, which would allow information-seekers to pose a question and get the program to retrieve the answer from a vast trove of data, and the many new insights that such dexterity could quickly and straightforwardly reveal. Carrillo cited a raft of performance analytics that will presumably become much easier to obtain, while Fisher envisioned how AI could support investors’ pursuit of value growth.
“We’re really focused on drivers of performance: sector allocations; market allocations; property selection; our valuation metrics and how they’re moving against the benchmark; impacts of leverage; and, at the end of the day, how our portfolio performed versus the underwriting and why,” Carrillo affirmed.
“I am definitely interested in arbitrage opportunities, looking at what a property is selling for versus what the model says the value should be, and taking advantage of AI’s scale to start layering in other data,” Fisher mused. “I think there’s a lot of interesting scale use-cases, which we’ll see embedded into other platforms to make investment decisions faster.”
Data is tapped to be increasingly vital once generative AI capabilities are in place, with bigger players better placed to reap the benefits.
“The more information that you’ve got, proprietary from your own investments or proprietary data from your own occupied portfolio, combined with purchased data, the more you’ll likely be able to make more intelligent decisions that are predictive and prescriptive,” Shtofman submitted. “I think the biggest firms with the biggest budgets for purchasing data will end up with an outsized advantage.”
