Staff ask questions in the chat widget — how many open orders over 30 days? which products sat unsold last month? — and an Aivell writes the query, runs it read-only against your database, and answers in plain words with the numbers. No report request, no waiting for the one colleague who knows SQL.
starts when: You type a question into the chat widget
In most companies, the data is plentiful and the answers are scarce. Between a question — “how many open orders are older than 30 days?” — and its answer stands the one colleague who knows SQL, a ticket queue, and a delay measured in days. So people don’t ask. They estimate, they reuse last quarter’s number, or they decide on gut feeling while the truth sits three tables away.
The irony is that the questions are almost always simple. They don’t need a data team; they need someone patient who knows the schema and has ten seconds. That colleague just never existed — until now.
The guardrails make this safe enough to hand to everyone: the connection cannot write, the sensitive schemas aren’t connected at all, and the log shows who asked what. Your database finally answers the phone — without ever leaving the building.
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:
when a question arrives in the chat, work out which tables answer it
write the query, run it read-only, sanity-check the result
answer in plain words, with the query shown underneath for the curious
round numbers sensibly; give exact figures on request
questions touching salaries or personal data — decline, point to Sofia
ambiguous question? ask back instead of guessing
keep the thread open for follow-ups like "and by region?"
every question and query goes in the audit trail
Connected with single-click passkeys — no copied tokens, no OAuth hell.
Every tool can be limited to exactly what this chore needs. For this one, you might set:
No. They ask the way they'd ask a colleague — "which customers went quiet this quarter?" — and the Aivell finds the tables. It learned your schema once, at setup, and keeps a note of your quirks, like which status codes mean "open".
No. Read-only is a guardrail enforced on the database tool itself, not a polite instruction. However a question is phrased, a write physically cannot run — and every query lands in the audit trail regardless.
No. The question, the schema, the query and the result all stay on the box. The model that writes the SQL runs onboard, inside your network — your data structure is as private as your data.
It asks back. "Sales this year — orders placed, or orders shipped?" One clarifying question beats a confidently wrong number, and the chore says so explicitly.
Your Aivell queries the database, builds the weekly report and mails it out on schedule — computed onboard, so your numbers stay behind your firewall.
An on-premise AI colleague watches your database and speaks up only when a threshold breaks — with the numbers, the trend and the context attached.
Every manual, contract and memo indexed on a box in your office. Your team asks in plain language and gets cited answers — nothing leaves the building.
An on-premise AI colleague for manufacturing back offices: turns order emails into system entries, works supplier portals, watches stock, reads shop-floor photos.
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.
onboard inference · air-gapped · fixed monthly fee