home / use cases / email and inbox / shared-inbox-triage

Shared inbox triage that never sleeps

An Aivell reads every message that lands in your shared mailbox, labels it, routes it to the right person, and answers the routine ones itself — from your own knowledge base, in the sender's language. It runs behind your firewall, so customer mail is never processed on someone else's servers.

Watch the demo

starts when: A message arrives in your shared info@ or support@ mailbox

The problem

The shared mailbox is where response times go to die. The average office worker already receives 121 emails a day; a shared info@ or support@ address multiplies that and removes the owner. Everyone can see it, so nobody handles it. Orders sit next to complaints, invoices next to spam, and the one message with a deadline is discovered three days late.

The usual fixes — rota schedules, color-coded flags, “whoever opens it owns it” — all rely on a human reading everything first. That first read is the bottleneck, and it is exactly the kind of repetitive judgment work that eats a quarter of the workweek.

How your Aivell does it

Your Aivell becomes the first reader — the colleague who opens every message the moment it arrives, at 2 PM or 2 AM:

  1. Every message is read on arrival. The trigger fires per email; nothing waits for a human to open the mailbox.
  2. Classified and routed by your rules. The chore holds your routing table in plain language: what counts as an order, who owns complaints, what must never be auto-answered. Changing the rules is editing a sentence, not reconfiguring a ticketing system.
  3. The routine ones get answered. Opening hours, delivery status, price list, “please resend the invoice” — answered immediately from your own knowledge base, in the sender’s language, with your tone. Sources only: a guardrail keeps answers inside your approved documents.
  4. The sensitive ones get escalated. Angry customers, legal wording, anything ambiguous — forwarded to the right person with a one-line summary, flagged urgent when the chore says so.
  5. You keep the overview. Every action lands in a triage log, and a morning digest tells you what arrived overnight, what was answered, and what is waiting on a human.

The mailbox stays yours, on your mail server — Gmail or plain IMAP. What changes is that it is never unread, and that the reading happens inside your office, not on an AI vendor’s cloud.

The chore, as you'd write it.

No flowcharts, no code — a chore is just a message to your Aivell, in your own words. It shapes it into a solid, guarded procedure and follows it to the letter. This one:

Giulia Aivell · triage the info mailbox ● active

read every new message in info@; skip newsletters and spam

label it — order, invoice, complaint, or other

orders go to sales@, invoices to accounting@, with a one-line summary

complaints — reply within the hour, apologize, and alert Marco directly

if it's a routine question, answer from the knowledge base, in the sender's language

not sure, or a legal threat? don't answer — forward to me, flag urgent

log every message and what was done in the triage sheet

at 8:30 send me the overnight digest — what came in, what was handled

You
Got it — I'll start right away, and check with you whenever something needs a human.
Giulia Aivell

The tools it uses.

Connected with single-click passkeys — no copied tokens, no OAuth hell.

Email reads, labels, routes and replies
Knowledge Base answers from your own documents, onboard
Spreadsheets keeps the triage log for the morning review

Sensible guardrails.

Every tool can be limited to exactly what this chore needs. For this one, you might set:

Email May reply only to the message's own thread; new outbound mail is limited to your team's addresses.
Email Deleting is off-limits — messages can only be labeled and moved, never destroyed.
Knowledge Base Answers may cite only your approved documents folder; when no source matches, escalate instead of improvising.

Questions, answered.

How does it know who should handle each email? +

You tell it — in the chore, in plain language. "Invoices go to Anna, complaints to Marco, legal threats straight to me, cc the lawyer." The Aivell follows the routing rules you wrote, and you refine them the same way you'd brief a new hire.

Will it answer customers with made-up information? +

No. It answers only from the knowledge base you gave it — your price list, your policies, your manuals — and a guardrail keeps it there. When nothing in your documents answers the question, the chore says escalate, so a human gets it.

Does customer email get processed in the cloud? +

No. The model runs onboard the box, inside your network. The mail stays on your mail server; the reading, labeling and drafting happen behind your firewall. Nothing is shared with any AI provider.

What if it classifies something wrong? +

Every action is labeled and logged, nothing is deleted, and the morning digest shows you what went where. Correct it once — "messages about refunds are complaints, not billing" — and the chore is updated like any procedure.

Related.

What is an Aivell?

An Aivell is an on-premise AI colleague: a small box that plugs into your network, runs its own AI onboard and takes the repetitive work off your desk. Prompts, documents, data — nothing ever leaves your office. Unplug the internet: it keeps working.

You describe each task as a chore, in plain language. Your Aivell turns it into a solid, guarded procedure and runs it in the background — with guardrails on every tool, a complete audit trail, and approvals in your hand when you want them. One-time setup, fixed monthly fee. No tokens, no overages.

Watch the demo

onboard inference · air-gapped · fixed monthly fee