According to the 2026 "AI Insights" survey conducted by IREM in partnership with AppFolio — based on responses from 3,662 real estate management professionals — AI usage across the industry has reached 58%, up from just 21% in 2023. That is nearly a tripling of adoption in three years. The same survey found something far less reassuring sitting right next to that number: most companies still have no written AI policy, and most professionals have never received formal training on the tools they are now using every day.
That gap — fast adoption, slow governance — is the defining operational risk in multifamily right now. Not because AI itself is dangerous, but because decisions that used to run through a person with judgment, context, and accountability are increasingly running through software with none of those things built in by default.
What "AI governance" actually means
AI governance is not a tool, a dashboard, or a compliance checkbox. It is the organizational structure that answers four questions before an AI system ever touches a resident, an application, or a maintenance ticket:
- Decision authority — which decisions can AI make unsupervised, which require human sign-off, and which AI should never touch at all.
- Policy — the written rules that define how AI tools may be used, by whom, and under what conditions.
- Human oversight — the specific points in a workflow where a person reviews, approves, or can override an AI-generated outcome.
- Accountability — who is responsible when an AI-assisted decision goes wrong, and how that responsibility is documented.
This is not a framework Sharpe Group invented. It mirrors the structure the National Institute of Standards and Technology (NIST) lays out in its AI Risk Management Framework, where the foundational "Govern" function establishes the policies, decision rights, and accountability structures that every other risk-management activity depends on. NIST is explicit that this framework applies not only to organizations building their own AI, but to any organization deploying third-party AI products or buying AI-enabled software — which describes the overwhelming majority of multifamily operators, who are licensing leasing chatbots, maintenance triage tools, and pricing software rather than building AI in-house.
Why "we use AI tools" is not the same as "we have AI governance"
Most operators already have AI somewhere in the stack — a leasing chatbot, an after-hours maintenance assistant, a pricing recommendation engine. Having the tool is not the same as having governed its use. Governance means someone can answer, in writing, questions like:
- If our leasing AI denies an applicant, can we show exactly why, and who reviews that decision?
- If a maintenance AI escalates — or fails to escalate — a safety issue, what was the rule it was following, and who set that rule?
- If a resident disputes an AI-influenced decision, do we have a documented trail of human oversight, or just a system log?
Without a governance framework, the honest answer to most of these is "it depends who you ask" — which is itself the risk. Inconsistent, undocumented decision-making is exactly the pattern that regulators, plaintiffs' attorneys, and investigative reporters look for.
The four pillars, applied
Decision authority has to be mapped property-by-property and decision-by-decision — a regional manager's authority to override pricing AI should not be identical to a leasing agent's authority to override a chatbot's response. Blanket authority is really an absence of authority.
Policy needs to be a living document, not a one-time memo. As new AI tools get adopted — and they will, continuously — the policy has to extend to cover them before they go live, not after an incident forces the question.
Human oversight should scale with stakes. A practitioner-level reading of the NIST framework puts this well: match the intensity of review to the risk level of the decision. A chatbot suggesting tour times needs less oversight than a system influencing whether an application gets approved.
Accountability is the pillar most organizations skip, because it is the least comfortable. It means naming, in advance, who owns the outcome of an AI-assisted decision — not discovering the answer for the first time during a complaint, an audit, or a deposition.
Where this leaves multifamily operators
The 58%-adoption, no-written-policy gap that IREM and AppFolio identified will not close itself, and it will not stay invisible. Every quarter that passes without a governance structure is another quarter of AI-assisted decisions being made without a documented, defensible rationale behind them.
This is the exact gap Sharpe Group's Governance Manual and Execution Modules are built to close — not by replacing the AI tools already in use, but by giving the organization the decision-authority structure, policy, and accountability trail that should sit above them.