Every email hitting the brokerage's shared inbox is read by an AI agent that classifies it — claim, endorsement, billing, carrier document — routes it to the right queue with a drafted reply, and files attachments to the correct client record automatically.
Typical results
92%
Auto-routed correctly
3 hrs/day
Sorting time saved
250+
Emails triaged daily
Figures reflect typical results for this kind of automation.
The challenge
The shared inbox took in 250-plus emails a day — claims, endorsement requests, invoices, carrier paperwork — all mixed together. Two back-office staff spent every morning sorting instead of processing, urgent claims sat behind newsletters, attachments were saved to the wrong client folders, and by afternoon the backlog had already rebuilt itself.
Before and after
Time from arrival to routed
−99%Emails still unsorted at noon
−100%Misfiled attachments
−88%The automation
We built a Make scenario with an OpenAI classification layer that reads every incoming email, identifies the category — claim, endorsement, billing, or carrier document — and matches it to the client record in the management system. Each message lands in the right team queue with a context-aware drafted reply ready to edit and send, attachments are renamed and filed automatically, and anything the model isn't confident about escalates to a human triage view.
How it works
- 01
Email lands
Every message hitting the shared inbox triggers the Make scenario the moment it arrives.
- 02
AI classifies
The email is categorized — claim, endorsement, billing, or carrier document — and matched to its client record.
- 03
Routed with a draft
The message lands in the right team queue with a context-aware reply already drafted for the handler.
- 04
Attachments filed
Documents are renamed by convention and saved to the correct client record in the management system.
- 05
Exceptions escalate
Anything the model isn't confident about goes to a human triage view instead of being guessed.
Starts from
- Shared inbox
- Carrier documents
Delivers
- Routed team queues
- Drafted replies
- Filed to client record
The impact
The team now clears the inbox by 10am instead of fighting it all day. 92% of emails route correctly without a human touch, drafted replies cut response time from hours to minutes, misfiled attachments nearly vanished, and the two sorters became two processors handling actual endorsements and claims.
Questions about this automation
What categories does the AI sort emails into?
The base setup sorts every email into claim, endorsement, billing or carrier document, and matches it to the right client record in the management system. Categories can be added to match how the brokerage's teams are organized, and each category maps to a specific team queue.
Are drafted replies sent automatically?
No. Each routed email arrives in the team queue with a context-aware reply already drafted, and the handler edits and sends it. The draft draws on the email content and the client record, which cuts writing time while a person still decides what goes out to the client or carrier.
What happens when the AI is unsure about an email?
Emails the model can't classify or match with confidence go to a human triage view instead of being routed on a guess. A team member sets the category and the client, and the email continues through the normal flow, with its attachments renamed by convention and filed to the right record.