home / use cases / email and inbox / email-reply-drafts

Reply drafts ready before you open your inbox

An Aivell reads each incoming email and writes a ready-to-send draft reply in your tone, pulling facts from your own knowledge base. Nothing goes out on its own — drafts wait in the thread for a human to approve, edit or discard. All the reading and writing happens onboard, behind your firewall.

Watch the demo

starts when: An email arrives in your mailbox

The problem

Answering email is a writing job nobody was hired for. Each message asks the same small labor: recall the context, dig up the fact, pick the words, get the tone right. Multiply that by the day’s stack and you arrive at 28% of the workweek spent inside the inbox — most of it composing slight variations of replies you have written a hundred times.

The usual fixes fail in opposite directions. Canned templates read like canned templates, and the sender notices. Full auto-responders go further and send things you would never say, to people who matter. What you actually want sits in between: a colleague who writes the first draft of everything — and always gives you the last word.

How your Aivell does it

  1. Every email is read on arrival. The trigger fires per message, day and night. By the time you open the inbox, the thinking is done.
  2. A draft is waiting in the thread. Written in your tone as briefed in the chore, addressed correctly, in the sender’s language. You open the thread and the reply is already there — Gmail shows it exactly where you’d start typing.
  3. Facts come from your documents, not from imagination. Prices, terms, delivery times are pulled from your knowledge base. Where no source exists, the draft carries a gap marker — one blank for you, instead of a blank page.
  4. The delicate ones are treated as delicate. Quotes get flagged for your numbers, heated messages get a holding acknowledgment and nothing else, and whatever routing you wrote — “German goes past Giulia” — is followed literally.
  5. You press send. Or edit, or delete. A guardrail keeps the Email tool physically unable to send, so the worst possible outcome is a draft you didn’t like.

The chore is eight lines of plain language, and every one of them is yours to change. Tighten the tone, add an exception, promote a correspondent to “always ask first” — it’s editing a briefing, not configuring software.

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 · draft replies for my inbox ● active

read each new email in my inbox; skip newsletters, receipts and automated mail

write a draft reply in my tone — short, direct, no corporate filler

pull facts from the knowledge base; if a fact isn't there, leave a [gap] marker instead of guessing

reply in the sender's language; German drafts go past Giulia before me

asking for a price or a commitment? draft it, but flag it "needs your numbers"

angry or legal-sounding mail — draft a short acknowledgment only, nothing more

save every draft inside its thread; never send anything yourself

at 17:30 list the drafts still waiting, so nothing sleeps in the outbox

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 incoming mail, writes draft replies
Knowledge Base supplies facts from your own documents

Sensible guardrails.

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

Email Can create drafts inside existing threads. Sending is blocked — only a human presses send.
Knowledge Base Facts may come only from the approved documents folder; when nothing matches, the draft gets a gap marker, not an invention.

Questions, answered.

Will it ever send an email without me? +

It can't. Sending is disabled by a guardrail at the tool level — not a polite instruction, a limit the box enforces. Your Aivell writes drafts; the send button stays human.

How does it learn to write like me? +

You brief it, in the chore, the way you'd brief an assistant — "short, direct, sign with first name only". It also sees the earlier messages in each thread, so it picks up how you already talk to that person. When a draft sounds off, correct the chore in one sentence.

Is my correspondence used to train some model? +

No. The model runs onboard the box in your office and learns nothing from your mail. Messages stay on your mail server; the drafting happens behind your firewall and is never shared with an AI provider.

What happens when it doesn't know the answer? +

The chore says don't guess. The draft arrives with a visible gap marker where the missing fact goes — a price, a date, a yes or no — so you fill in one blank instead of writing the whole reply.

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