Ask in the dashboard header opens a panel that answers questions about your
own helpdesk in plain language: who is at capacity right now?, what docs are
we missing?, why did first reply time drop?
It is an analytics assistant, not a support agent. It reads; it does not act.
What it can look at #
It has six read-only ways of looking at your site:
| It can fetch | Meaning |
|---|---|
| Helpdesk overview | Headline figures: tickets by state, doc counts, sentiment, service levels. |
| A list of tickets | Across every client, newest first, filtered by status or agent. |
| One ticket | With its full conversation. |
| Agent stats | Per-agent load, first-reply and resolve medians, at-capacity flag. |
| Doc gaps | Searches that returned nothing, busiest first. |
| A list of docs | Published docs, newest first, with an optional search. |
It works over 7, 30 or 90 days, and lists come back 20 rows at a time.
The panel shows which of these it used, under “How I got this.” An answer is
auditable rather than asserted.
What it deliberately cannot do #
- It cannot message a client. There is no way for it to reply to a ticket,
and it is told outright never to offer. - It cannot read anything outside that list. It has no free run of your
database, so it never touches private posts, passwords or your other plugins’
data. - It cannot change anything. No status changes, no reassignment, nothing.
That waits until the panel can ask you to confirm first — a status change
emails your client, and no email should ever go out because the assistant
guessed wrong. - It cannot invent a figure. It is required to cite ticket numbers and
forbidden from stating a number that did not come back from a real lookup. - It has no opinion on tone. Per-ticket sentiment does not exist, so it does
not guess at one.
How long an answer takes #
Complex questions take several lookups. The panel gives itself about 70 seconds
and at most four rounds of looking things up. If it runs out, it tells you what
it has rather than starting something it cannot finish.
Conversations #
Each question continues the current conversation. New question starts a
fresh one. History is kept for two hours and up to 30 messages, then
discarded — nothing about your conversations is stored permanently.
Requirements #
The site has to be connected to the AI service. It does not need the tickets
module, but with no tickets most of the answers will be thin.
Starter questions #
The empty state offers one per area, and Browse all questions lists the rest:
- Tickets — Why did first reply time drop 46%?
- Agents — Who is at capacity right now?
- Chatbot — Why are 45% of sessions abandoned?
- Docs — What docs are we missing?
Was this doc helpful?