home / use cases / knowledge and answers / self-writing-procedures

Chores that write themselves

Describe a task to your Aivell the way you would brief a new colleague. It asks until the goal is clear, drafts the chore in plain language, proposes guardrails and a schedule, and hands you the draft. Nothing runs until you approve — and the draft stays readable enough to check in a minute.

Watch the demo

starts when: You describe a task to your Aivell in the chat

The problem

Everyone has a list of things that should be automated. Almost none of it happens, because the person who knows the procedure — how orders really get checked, who really approves what — is never the person who can build the automation. The knowledge sits with the office manager; the tooling sits with a consultant who costs money and needs a specification. Between them, 62% of employee time keeps going to repetitive work that everyone agrees a machine should do.

The specification is the real killer. Writing down a procedure precisely enough to automate it is itself a skilled, tedious job — so the backlog grows and the retyping continues.

How your Aivell does it

  1. You brief it like a new hire. In chat, in your own words: “every Monday, chase the suppliers who haven’t confirmed delivery dates.” No forms, no flowchart tool.
  2. It interviews you. Which suppliers? Chase how — email, and in what tone? What counts as confirmed? The Aivell asks until the fuzzy parts are gone, and checks the knowledge base so it extends an existing chore instead of cloning it.
  3. The draft comes back readable. Numbered plain-language steps, a proposed guardrail for every tool involved, a suggested schedule — and explicit flags on anything that would contact the outside world or touch money.
  4. You approve; only then does it exist. The draft lands with Luca. Until he says yes, the chore is text. A guardrail makes activation a human-only act — the Aivell cannot switch on what it wrote.
  5. The first week comes with a report. Once live, the new chore accounts for itself: what ran, what it did, what it escalated. You tighten a sentence or two, the way you’d correct any new colleague.

The procedure that lived in someone’s head is now written down, reviewed, guarded and running — and writing it down cost a conversation, not a project.

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 a chore from a briefing ● active

listen to the task description; keep asking until the goal is unambiguous

check existing chores first — extend one rather than duplicate it

draft the steps in plain language, short enough to read in a minute

propose a guardrail for every tool the new chore would touch

flag anything that mails outside the company or moves money

suggest a trigger or schedule, but leave it switched off

send the draft to Luca; nothing runs before his yes

after the first week live, report what the new chore actually did

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.

Chat widget takes the briefing, asks follow-ups
Knowledge Base checks existing chores and procedures
Schedule proposes when it should run
Email delivers the draft for approval

Sensible guardrails.

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

Schedule Drafts are saved inactive. A chore cannot switch itself — or any other chore — on; activation is a human action.
Email Draft approval requests go to the owner named in the briefing, nobody else.
Chat widget A drafted chore may only use tools the person briefing it already has access to — no quiet privilege escalation.

Questions, answered.

Will it start doing the task while we're still talking about it? +

No. The conversation produces a draft, and a draft is inert by design — a guardrail keeps every new chore switched off until a human activates it. Talking about mailing customers never mails a customer.

Our procedures are how we run the business. Where do they live? +

On the box, next to everything else. The chores it drafts, the conversations that produced them, the knowledge base it checked — all onboard, behind your firewall. Your operations manual doesn't become someone's cloud data.

Can it change a chore that already exists? +

It can propose a change — say "the invoice chore should also handle credit notes" and it drafts the amended version. The same rule applies — the current chore keeps running untouched until you approve the new text.

What if the draft gets a rule wrong? +

Then you edit a sentence — the draft is plain language, not configuration. And because the first week ends with a report of what the chore actually did, a wrong rule surfaces as a line in a summary, not a surprise in production.

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