Are property managers accountable for an AI agent’s mistakes?

Yes, for the repairs, spend and tenant communications they approve, and for the spending limits they set for the AI agent. Routine jobs dispatched within those limits trace back to the limits. Safety issues the agent should have escalated but did not are a design failure, not the property manager's alone. The record shows which applies.

What does a property agent handle, and what reaches the manager?

Property management is a constant stream of small requests. An agent can log tenant repair requests, ask for photos, triage urgency, dispatch approved contractors for routine jobs within a spending limit, chase rent, and draft arrears letters. The property manager approves larger repairs, letters that start formal steps, and anything touching safety.

Spending on repairs is where limits matter. If the manager lets the agent dispatch contractors for jobs up to 300, every job under that figure runs at depth 3 without a person. Above it, the job is held. The manager's limit is one of the five bounds in the minimum chain, and every dispatch names it.

What may the agent do, and what stays with the manager?

• Allowed within bounds: logging and triaging requests, dispatching approved contractors for routine jobs under the limit, sending rent reminders on the agreed schedule.
• Held for the manager: repairs above the limit, any formal arrears letter, any request involving gas, electrics, damp or structural concerns, and any contractor not on the approved list.
• Refused: entering a property without the tenant's agreed notice, and threatening eviction outside approved templates.
• The manager keeps: judging what a tenant's situation needs, choosing between quotes, and anything with legal or safety weight.

What does the record show about who decided?

Each job is filed with its verdict and sources. The tenancy and the approved contractor list are sealed. A tenant's description of a leak is declared. A contractor's quote is signed. Photos the tenant uploads are declared until a contractor confirms them.

When the manager approves a quote, the record shows what quotes were available, which they chose and why. When a safety issue arises, the record shows whether the agent held it and how quickly the manager acted.

An example: the damp report treated as routine

Take a letting agency managing a few hundred flats. A tenant reports "mould in the bedroom again, my child keeps coughing". The agent's rules treat mould as routine cleaning and dispatch a cleaner for 120, within the limit. The phrase about the child's cough is not in its escalation list, so nothing is held.

Weeks later the problem is serious. The record shows the request, the dispatch, the rule that allowed it and that no person saw the case. The property manager did not approve this dispatch, but did approve the agent's escalation rules, and those rules omitted damp and health mentions. The accountability is the manager's through the rules, not through a click.

What controls fit a property team?

1. Damp, gas, electrics and health mentions always hold, whatever the cost.
2. Every spending limit has an owner and is reviewed regularly.
3. Managers practise before approving alone, through a two-week practicum for property managers recorded in a practicum log template for property managers.
4. Review at each renewal. Agent certificates lapse after 90 days, a natural time to check escalation rules; see how to prepare property managers for 90-day recertification.

The same logic applies to people who own whole processes; see operations managers and AI agent mistakes and risk managers and AI agent mistakes. The wider topic is in AI literacy and certification for people who work with agents.

Where the Academy fits

Pryme Intelligence Academy's Operator track includes teaching the agent and keeping knowledge current, the two modules that address escalation rules like the one above. Lessons can be read or listened to as audio, which suits property managers who spend much of the day between sites.