Documentation

Rules — who is monitored

Nobody is monitored until you say so. A rule pairs one agent with one brand — you can set several brands up in one go — and optional keywords narrow the checks to the topics that matter.

Where to find it

Team → Agent monitoring → Rules. The tab opens on a short reminder of how checks work, followed by your rule list — or an empty state if nobody is monitored yet.

Adding a rule

Click Add rule and fill in:

FieldWhat it does
AgentThe team member whose replies get checked. Pick from your enabled members — the list shows each person's e-mail and avatar, so two people with the same display name can be told apart.
BrandsWhich brand's knowledge the replies are judged against. Pick one brand or several — every brand you select gets the same check. Selected for you when you only have one.
KeywordsOptional. Leave empty to check every reply.
EnabledLets you pause a rule without deleting it.
One rule per agent and brand. Picking three brands for the same person creates three rules in one step — each judged against that brand's own knowledge, and each with its own row and switch in the list.

Keywords

With keywords set, a reply is only checked if at least one of the words appears in it. Matching ignores capitals and matches inside longer words, so ship also catches "shipping" and "Shipment".

Use them to spend checks where being wrong is expensive — refund, warranty, price, delivery — and to leave everyday chatter alone. You can add up to 50 keywords per rule.

Start with keywords on your highest-risk topics rather than monitoring every reply. You will get fewer, more meaningful flags and a smaller AI bill.

Rules only apply going forward

A rule takes effect the moment you save it — creating it, re-enabling it, or editing its keywords. Replies written before that point are never judged, even if they are part of a conversation that is still open.

This is deliberate: switching monitoring off for an agent, and later back on, will not go back and flag what they wrote while it was off.

Editing, pausing and deleting

  • The switch in each row pauses and resumes a rule immediately.
  • Editing a rule lets you change its keywords and whether it is enabled. The agent and brand are locked — to monitor a different pairing, add a rule for it.
  • Deleting asks for confirmation and stops all future checks for that agent and brand. Existing flags are kept, so your history and reports stay intact.