Urgency read on arrival
Each inbound message is scored low, medium, high, or emergency. Emergency content is escalated rather than queued behind whatever arrived first.
Sentus Inbox
Email Copilot reads inbound mail, files it against the right case and work order, and drafts the reply — with a human confirming the facts before anything sends.
Approvals, vendor replies, owner questions, and habitability complaints all arrive as email. In most companies that mail sits in a shared inbox that no system can see, so the same request gets re-explained three times and the one message that mattered gets buried under forty that did not.
Email is not a side channel in property management. It is where approvals, disputes, and vendor coordination actually happen — and it is usually the one channel the software cannot see.
The goal is not to write email for you. The goal is to stop the mailbox from being a place where operating work goes to disappear.
Every item below is a shipped behavior in the Owner and Management portals, not a roadmap item. Copilot reads, files, and drafts — a person still decides what leaves the building.
Each inbound message is scored low, medium, high, or emergency. Emergency content is escalated rather than queued behind whatever arrived first.
Messages are labelled as maintenance request, rent payment, lease question, noise complaint, move-in, move-out, legal concern, emergency, or general inquiry — so the inbox can be worked by category instead of by scroll position.
Copilot proposes one of: create a work order, open a case, reply, escalate, or no action needed. That suggestion is what connects the mailbox to the rest of the workflow.
Legal threats, discrimination claims, habitability complaints, safety issues, and anything scored as an emergency are flagged for human review automatically. Copilot does not quietly answer those.
The message is linked to a case record and its thread, and the action is written to the activity log. The next person to open the case sees the email as part of the history, not as a forwarded screenshot.
Drafts are generated with the thread already in context, and the evidence the draft leaned on is shown next to the composer so the reviewer can check the source rather than trust the prose.
Tone lock holds the voice steady across drafts and across the people sending them, so the third reply in a dispute does not read like a different company than the first.
Before a draft goes out, Copilot surfaces claims in the text that the attached evidence does not support. It is a check against the specific failure mode of AI drafting: fluent sentences that promise something nobody verified.
When a draft is rewritten, the before and after are shown side by side, so the reviewer can see exactly what the rewrite changed instead of re-reading the whole message.
The send step asks the person to confirm they verified facts, evidence references, lease language, and timeline commitments. Accountability for what goes out stays with a named human.
A manager can speak the reply instead of typing it. The memo is transcribed and becomes the starting draft — useful when the answer is long and the person is standing in a hallway.
A drafted reply can go out as a text instead, so the answer reaches the resident on the channel they actually read without leaving the record behind.
Straight answers about what sends automatically, what does not, and where the mail ends up.
No. Copilot classifies the incoming message and drafts a response, but sending requires a person to confirm they verified the facts, the evidence references, the lease language, and any timeline commitments in the draft. The accountability for what goes out stays with a named human.
Those are flagged for human review automatically. Legal concerns, discrimination claims, safety issues, and anything classified as an emergency are routed for a person to handle rather than answered as routine correspondence.
No. Email Copilot works against the Google or Microsoft mailbox your team already uses. Residents, owners, and vendors keep emailing the same address they always have.
When Copilot classifies a message as a maintenance request, the suggested action is to create a work order. A person confirms it, and the work order carries the email thread and case link with it — so the maintenance record and the correspondence stay attached instead of living in two systems.
Two ways. The evidence rail shows the reviewer what the draft actually drew from, and the claim coverage check flags statements in the draft that the attached evidence does not support — before the send step, not after. Neither replaces a human reading the message.
It runs in the Owner portal and the Management portal. Both work from the same case and thread records, so a message handled by one is visible to the other rather than siloed per role.