0seat
Knowledge base

Drawn from your conversations, approved by you

You do not write a knowledge base: you review the one your resolved conversations already contain. Every night, 0seat extracts candidate entries from them; your team decides what gets in.

Create an account Request a demo

The problem

The answers exist, scattered across thousands of closed conversations. Writing them by hand takes weeks, and the base goes stale the next day. Meanwhile, the same questions keep coming.

How 0seat handles it

Nightly extraction

Every night, resolved conversations are reviewed to extract generalised question-and-answer pairs, personal data masked, with the original conversation as the source.

Mandatory human approval

Every entry waits in a queue. A team member approves, edits or rejects it. No entry is used before that.

Your canned responses as a starting point

Your inbox’s predefined replies come in as candidates, to be reviewed like any other.

0seat console, Knowledge base tab: 12 entries to review, 148 approved, one entry “My domain name stopped responding this morning” with its answer and the Approve, Save without deciding and Reject buttons.
Console screenshot: the queue of candidate entries, with approve, edit or reject.

What you control

  • Who approves: the owner, administrators and agents assigned to the space.
  • The text of every entry, editable before approval and after.
  • Removing an entry at any time: it stops being used immediately.
  • Running extraction manually from the space’s health page.

Honest limits

  • Extraction only reads your inbox’s resolved conversations, not your public pages or PDFs.
  • Full-text search is configured for French: a knowledge base in another language is not supported yet.
  • 0seat does not yet flag questions with no answer in the base.

Frequently asked questions

How long until the first entries?

One night after connecting, or a few minutes with a manual run. The number of entries depends on your resolved history.

Do entries contain customer data?

Personal data is masked before extraction, and entries are generalised. You review every entry before it is used.

Can an approved entry be edited?

Yes, at any time from the console. The edited version is used as soon as it is saved.

What happens to a rejected entry?

It is never used. The rejection is kept; if a similar answer comes back in the history, it goes before a reviewer again, flagged as close to a rejected entry.

Your team stays in charge. 0seat does the groundwork.

Connect an inbox, approve the suggested knowledge base, and get your first prepared replies.

Create an account Request a demo

See also

Reply drafts

The draft arrives as a private note with the knowledge base entries behind it. The agent sends, edits or discards it. Below the threshold, no reply.

Human handoff

Money, disputes, cancellations, a request for a person, a second follow-up, an answer missing from the base: an agent takes over, whatever the mode.

Measured costs

The real cost of every model call, per conversation, in the console. A monthly cap per space, checked before each call.