The answer depends on which building
Two properties have different rules for the same question. An answer that is correct for one is wrong for the other, and the difference is easy to miss under time pressure.
For property management companies with 10 to 50 people
Tenants and owners ask what is allowed, who pays and how long it takes. The answer sits in a lease, a building rule or a service agreement, and it differs by property, so whoever answers has to find the right document first.
A grounded agent answers from the documents for that property, cites them, and refuses when they do not settle the question. It should never average across properties to produce a plausible answer.
Book a readiness assessmentTwo properties have different rules for the same question. An answer that is correct for one is wrong for the other, and the difference is easy to miss under time pressure.
Maintenance and access requests come by phone, email and text. They are unstructured at the point they occur, so they get retyped into a system later, or not at all.
An expense or an access request was agreed somewhere in a thread. Reconstructing who approved it, and when, means reading back through messages.
We review the Microsoft 365 tenant, licensing, a representative set of leases and building rules, and the requests that arrive most often. You get a written recommendation on where a grounded agent could help, what would block it, and what should stay with a person.
Fixed scope and fixed fee. The report can recommend not building anything yet.
One grounded agent in production over the agreed document set, scoped so that an answer is tied to a specific property. Cross-property confusion and unsupported questions are part of the acceptance checks.
The build scope and delivery window are agreed after the assessment.
Add scoped automation for the intake side, so a request becomes a structured record with an owner and an approval trail instead of a message in a thread. Document who owns each property’s documents and how changes are published.
Each extension is scoped after the assessment, with its delivery window agreed before work starts.
Staff get cited answers from your policy library instead of asking HR the same questions again.
Requests are captured in Teams, logged in Dataverse and routed for approval without email chains.
Incoming documents are classified and routed, with uncertain cases sent to a person to review.
Scope is the whole design. The agent has to establish which property a question is about before it answers, and the acceptance checks include questions that could plausibly apply to more than one. If it cannot tell, it should ask or refuse.
That is a decision to make deliberately, not by default. Who can reach an agent determines which documents it may draw on, and the assessment checks whether a tenant-facing audience is supportable before anyone builds for it.
Those stay with a person, and the automation starts once the request is written down. We would rather make the written path fast and reliable than pretend a phone call can be captured automatically.
A review of what is worth building on your tenant and what isn’t. There is no obligation to proceed to a build.