home / use cases / knowledge and answers / auto-kb-ingestion

A knowledge base that feeds itself

An Aivell watches a mailbox and your network folders for new documents, reads each one onboard — office files, PDFs, scans, photos — gives it a clean title, files it, and indexes it into the knowledge base so search finds it minutes after it arrives. No uploads, no manual tagging, no stale handbook.

Watch the demo

starts when: A new document lands in the docs mailbox or a watched folder

The problem

Every knowledge base is a snapshot of the week someone had time to maintain it. The launch is glorious: everything tagged, everything findable. Six months later the handbook is three versions behind, the newest spec sheet lives in someone’s inbox, and the search returns the 2022 price list with total confidence.

The rot isn’t laziness — it’s that keeping documentation current is a job nobody was hired for. Someone has to notice the new file, convert it, name it sensibly, put it in the right place and update the index. That’s classic invisible admin, the kind that quietly costs businesses around $17,000 per employee per year. So it doesn’t happen, and the knowledge base becomes a museum.

How your Aivell does it

  1. New documents announce themselves. Anything mailed to the docs address or dropped into a watched folder triggers the chore. Nobody has to remember a second step — saving the file is the whole process.
  2. Everything is read on the box. PDFs, office files, scans, photos of whiteboards and paper forms — the Documents tool and the onboard vision model read them behind your firewall. No cloud conversion service, no upload.
  3. Filing follows your convention. The Aivell titles the document cleanly, puts it under the right section, and replaces superseded versions — archiving the old ones rather than destroying them.
  4. Sensitive material waits for a human. The chore names the exceptions: anything touching personnel pauses until Paolo says yes. What should never be indexed simply lives outside the watched folders, where guardrails keep the Aivell from looking.
  5. Search catches up in minutes. Once indexed, the document is findable by everyone using the knowledge assistant — and the monthly review sweep flags what has grown old enough to question.

The result is a knowledge base with the one property none of them have: it is current on a random Tuesday, not just the week after the cleanup.

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 · keep the knowledge base fresh ● active

watch docs@ and the handbook folder on the share for new files

read each file onboard — office files, PDFs, scans, whiteboard photos

give it a clean title and file it under the right section

newer version of an existing document? replace it, archive the old one

anything mentioning salaries or personnel waits for Paolo before indexing

index it into the knowledge base — searchable within minutes

reply to the sender saying what was indexed and where it lives

first monday of the month, list documents older than two years for review

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 watches the docs mailbox, confirms indexing
Network folders watches drop folders, files the originals
Documents & Vision reads every format onboard
Knowledge Base indexes it so search finds it

Sensible guardrails.

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

Network folders Reads from the drop folders; writes only into the handbook and archive folders. The rest of the share is read-only.
Knowledge Base Can add and update entries. Removing a document from the index requires a human.
Email Replies go to the original sender only, and only on your own domain.

Questions, answered.

What can it actually read? +

If a person can read it, so can the box — office documents, PDFs, scanned contracts, a phone photo of a whiteboard from this morning's meeting. The vision model runs onboard, so a photographed page gets the same privacy as a typed one.

Does indexing happen on some vendor's servers? +

No. Conversion, reading and indexing all run on the box inside your network. A draft contract dropped in the folder is searchable ten minutes later and has never crossed your firewall.

What about documents that shouldn't be searchable by everyone? +

Two layers. The chore routes sensitive material — the example above holds anything about personnel for Paolo's approval — and the guardrails limit which folders the Aivell can index at all. What you never point it at, it never sees.

What happens to superseded versions? +

The new version replaces the old one in the index, and the old file moves to the archive rather than being deleted. Answers cite the current version; the paper trail keeps every previous one.

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