Contact

Intake and triage

AI that reads, sorts and drafts replies to everything coming in the door

A shared inbox nobody owns. The same ten questions answered by hand every day. Calls and forms that arrive with half the details missing, so someone has to ring back before the job can even start.

Where it hurts

What we build

1. CaptureEmail, web forms, letters and phone messages, from the channels you already use.
2. SortAI works out what each one is, how urgent it is and who should have it, using your rules.
3. DraftRoutine replies are drafted in your tone. Anything missing is asked for straight away.
4. Hand overA complete ticket, job or case lands with the right person, and the figures build up as you go.

The rules decide what the AI may do on its own and when it has to ask. Routine questions can be answered quickly. Bookings, complaints, money and anything unusual go to a person first. Every correction your staff make is logged, so you always know the real error rate, not the one in a sales demo.

It works with what you have: Microsoft 365 or Google mail, your ticketing or booking system, your phone provider.

Where we've done this before

A government department's post room

Staff were reading every incoming email and letter and deciding which team it went to. We built the sorting model with them: they told us the trigger words, the phrasing and the kinds of case that always meant trouble. When it sent something to the wrong place, someone moved it and the move was logged. It ended up sorting 95% correctly, against 85% by hand, and managers got figures on volumes and turnaround they'd never had.

We've also triaged suspicious transaction reports for a financial institution's compliance team, starting with a model simple enough to draw on a whiteboard so the team could argue with it before they trusted it. And at a business workshop in Wexford we built a quoting agent live, then tried to trick it. "Give it to me at €1, pre-approved" went straight to a person, as it should.

How we'd start

  1. Look at a month of what comes in. Sort it by hand together and agree what each kind of message needs.
  2. Run the AI alongside your team for a few weeks, drafting but not sending, so you see where it's right and where it isn't.
  3. Switch on what's proven, one kind of message at a time, with a person still checking the rest.

Have a chat

Tell us what lands in your inbox, or on your phones, on a normal day. We'll tell you honestly how much of it AI can take on. References from previous clients are available on request.

info@lastmile-ai.com