How to Document an ESA Denial So the HOA Stays Protected

What changed for ESA requests in 2026

Florida's 2026 Condo Owner's Bill of Rights reinforced assistance-animal protections that already sat inside federal fair-housing law, and it raised the cost of getting the process wrong. An emotional support animal is not a pet. Under fair-housing rules, it is a reasonable accommodation request, which means a board cannot apply its pet policy, its weight limits, or its breed restrictions to it the way it would to a household dog.

That distinction is where boards get hurt. Directors read "strengthened rights" as "we can never say no," panic, and either rubber-stamp every request or freeze and let requests age past a reasonable response window. Both reactions manufacture liability.

A denial is legal. Federal guidance allows boards to deny an assistance-animal request when the requester is not a person with a disability, when the animal poses a direct threat, or when the request would impose an undue burden. The problem is almost never the legal grounds. It is that the board cannot prove, months later, what it asked, what it received, and when.

Why the risk is the record, not the decision

The uncomfortable truth: most HOA ESA denials that turn into complaints were probably defensible on the merits. The board had grounds. What it did not have was a file. Six months later nobody remembers whether the resident's letter arrived before or after the denial, whether the board asked for reliable documentation or just ignored the request, or whether the identical situation last year got a faster answer.

Fair-housing enforcement leans heavily on inference. When an investigator cannot see a good-faith interactive process, they infer there was not one. Silence and delay read as avoidance. According to guidance summarized by resources like the Florida Department of Financial Services, consumer complaints frequently hinge on documented timeliness and consistency of communication.

This is exactly the kind of documented, deadline-driven work an AI operations agent absorbs well. It does not decide anything. It captures the request the moment it lands, opens a timestamped file, logs every document in and every response out, and flags when an interactive-process step is overdue. The board keeps the judgment. The agent guarantees the judgment leaves a trail.

The interactive-process steps an agent should track

The interactive process is the back-and-forth a housing provider must engage in when an accommodation is requested. Skipping it is the single most common fatal error. Here is the sequence a well-configured intake agent timestamps and routes, with the human decision points marked.

1. 01

   1. Capture the request and open the file (agent)

   The moment a request arrives by email, portal, phone, or letter, the agent logs the date, channel, resident name, unit, and exact wording. A request does not have to say "ESA" or "reasonable accommodation" to count. If a resident writes "my doctor says I need my dog for my anxiety," that is a request, and the clock starts. In One Home Agent deployments, Riley Resident captures this at first contact so nothing sits unacknowledged in an inbox.

2. 02

   2. Acknowledge promptly (agent drafts, human sends)

   A dated acknowledgment goes out fast confirming the request was received and describing what happens next. This single step defeats the most damaging inference: that the board ignored the resident. The agent drafts using approved language; a human approves before it sends.

3. 03

   3. Request reliable documentation only if the disability is not obvious (human decides scope)

   If the disability and need are not obvious, the board may request reliable supporting documentation from a professional in a position to know. The board may NOT demand medical records, a specific diagnosis, or a specific provider. The agent logs exactly what was requested and when, so the request itself is defensible.

4. 04

   4. Log what arrives and flag gaps (agent)

   When documentation comes in, the agent files it with a timestamp and flags whether it addresses the disability-nexus question or leaves it open. If a follow-up is needed, the agent notes the reason so a later reviewer sees the process was substantive, not a stall.

5. 05

   5. Route to the board or delegate with a decision packet (agent assembles, board decides)

   The agent assembles a clean packet: request wording, dates, documentation received, prior comparable decisions, and the applicable standard. The board or its counsel makes the call. The agent does not recommend approve or deny; it presents the record.

6. 06

   6. Deliver and timestamp the decision (human signs, agent logs)

   The written decision, whether approval, denial, or request for more information, is delivered and logged with the specific basis. A denial states the fair-housing ground relied on. The agent closes the timeline with the delivery date and stores the full file for the retention period.

What a defensible denial file contains vs. what sinks it

Notice that not one row in the defensible column is about the outcome. Every row is about process and proof. A board can deny an ESA request and win when the left column is filled in. A board can approve one and still get named in a complaint over the delay if the right column describes its file.

The most common sinker is the documentation request that quietly demands too much. Asking for the resident's full medical file, or insisting on a particular doctor, converts a lawful information request into an unlawful one. The agent's value here is narrow and real: it logs precisely what was asked, so if the request was overbroad, that is caught before it becomes evidence.

The selective-treatment trap across owners

This is where volume quietly builds risk. A single community might see a handful of ESA requests a year, handled by whoever was managing that week. Different manager, different mood, different response time, different documentation ask. Each decision may be reasonable alone. Lined up, they look arbitrary, and arbitrary is the language fair-housing complaints are written in.

An agent with institutional memory per community fixes the part humans are worst at: remembering what was done last time. When a new request lands, the agent surfaces prior comparable requests and how long each took. The board still decides case by case, but it decides against a visible baseline instead of a blank page. That is the difference between a defensible pattern and a discoverable one.

“Boards do not lose these cases because they denied someone. They lose because they cannot show the person next door was treated the same way. The agent's job is not to make the call. It is to make sure the call has a comparable next to it and a clock behind it.”

Todd Paton, Partner, One Home Agent

Where the human stays firmly in charge

The agent never decides an accommodation request. That line matters legally and practically. Fair-housing determinations require human judgment about disability nexus, direct threat, and undue burden, and they carry liability that no automated system should hold.

Everything above the decision line is judgment. Everything below it, intake, acknowledgment, timestamping, gap-flagging, packet assembly, consistency lookup, and retention, is documented busywork the board is currently doing badly by hand at 9pm. That split is the whole design. Approvals and denials both get a clean record, and the humans keep the calls that only humans should make.

Build the record before the complaint