What an Algorithm Can't See on a Property Tour

A 32-unit deal can look clean on a screen in the morning. Then the basement reveals a boiler nearing replacement, and the manager explains why several units have lingered vacant. The economics shift before the purchase price does.

Commercial real estate is already testing how much of the earlier work can move to software. The 2025 Global Real Estate Technology Survey by JLL found that 88% of investors, owners, and landlords had started piloting AI, with most pursuing several use cases at once. Only 5% reported achieving all of their program goals, a reminder that adoption and useful implementation are still different things.

For small multifamily, the appeal is practical. A machine can turn a messy rent roll and T-12 into a consistent first pass much faster than an analyst working line by line. The time saved is valuable because it lets a team screen more opportunities without giving every deal the same amount of attention.

Use AI Where the Work Repeats

Software earns its keep when the work follows a repeatable pattern. It can take an inconsistent rent roll or T-12 and put the same fields in the same places every time. An analyst can then test changed assumptions without rebuilding the file by hand. On a small deal, compressing that work makes diligence more economical and leaves more time for the variables that actually require judgment.

The constraint is the quality of the information going in. Deloitte, in its 2025 commercial real estate outlook, found that data readiness remained one of the main obstacles to scaling AI in real estate. Only 14% of respondents believed their organizations had well-structured data collection and management processes in place alongside robust privacy policies.

Small properties often make the problem more visible. Records may come from different owners or management systems, and local operating history can be thin. A fast model still needs a trail back to the source, especially when the rent roll, general ledger, and seller narrative do not line up neatly.

The Building Supplies Its Own Data

The tour can recast a low repair expense, revealing whether it reflects efficient operations or work that has merely been postponed. Strong occupancy can look different after learning how long turns are taking and how much effort management needs to keep residents in place.

Industry practice still treats physical observation as a core part of diligence. ASTM E2018-24, the standard guide for baseline property condition assessments, centers a current walk-through survey and pairs it with document review and interviews to identify material physical deficiencies. The standard itself recognizes that documents alone do not describe the full condition of a property.

Walking a small apartment building also shows how operations are translating into physical reality. A turn that looked routine in the budget may still be unfinished, and a worn common area can expose gaps in the capital plan. Conversations on site often explain why. Each finding becomes useful once it is carried back into the economics.

Scale makes those observations matter sooner. A turn that looked routine in the budget may still be unfinished, and a worn common area can expose gaps in the capital plan. One unexpected capital item can absorb a meaningful share of first-year cash flow, while a few slow turns can materially weaken occupancy. The model can price those effects accurately once someone spots them.

Let the Tour Rewrite the Model

The strongest process closes the loop between digital screening and physical diligence. Early automation should help a team reach a view faster, then the site visit should be allowed to challenge it. A model that survives the tour is more useful than one protected from inconvenient facts.

Suppose turn times are consistently running closer to a month than the model assumed. Vacancy and make-ready costs should move with that observation. A recurring moisture problem deserves the same treatment in the capital plan. Physical findings matter when they change an assumption, and the value of technology is partly measured by how quickly those changes can be reflected across the deal.

Local judgment becomes especially valuable when comparable data is sparse. Software can surface nearby transactions and operating benchmarks, yet it cannot fully explain why one block leases differently from another or whether a management problem is temporary. Those calls require context built from repeated exposure to the market.

At Gilberti Group, technology is useful when it buys more time for judgment. Faster screening helps us say no earlier and reserve deeper attention for the assets worth pursuing. Once we are on site, the property gets to challenge the spreadsheet.

To discuss how Gilberti Group evaluates small multifamily opportunities and property-level risk, schedule a conversation with our team.

Next
Next

Family Offices Are Going Direct. Small Multifamily Is Where It Gets Hard.