Asking HQ a question is the no-risk half of using it. Searches, lookups, and pipeline questions are reads — they don't change anything — so HQ just runs them and answers. There's no plan to approve and nothing to undo. Type a question, get an answer.

What counts as a question

Anything that looks something up rather than changes it. HQ runs these the moment you ask:

Find candidates. "Senior backend engineers in Berlin who applied this month." HQ interprets the sentence and searches — the same engine behind smart search, reachable from chat.
Pull up one record. "Show me Priya Nair's profile" or "what's the status on the Staff Designer role." HQ fetches the candidate or job and summarizes it.
Read the history on someone. Their notes, their feedback, the jobs they've applied to, the stage each application sits at.
Check the pipeline. "How many candidates are in interview across all open roles." HQ pulls the pipeline summary — counts by stage and by job.
Look at scheduling. This week's interviews, what's still unscheduled, the AI scheduling events waiting for review.

How an answer arrives

You type a question and send it.
HQ picks the right tool — looking up one named candidate is a different tool than searching a list — and runs it. You see each step it takes.
The answer streams in as HQ writes it, so you're reading before it's finished thinking.
When it's done, any candidate or job HQ named in the answer becomes a clickable link straight to that record.

Those links are the fast path. Ask "who's in the offer stage for the Backend role," and instead of a list of names you have a list of links — click one, you're on the profile.

A conversation remembers itself

A chat keeps its context. Ask HQ to find a candidate, then ask "what jobs has she applied to" — it knows who "she" is. You can narrow a search across several messages without repeating yourself.

This works best when one chat covers one task. A conversation that wanders from a candidate search into scheduling into a job edit gets long and unfocused, and HQ has more to wade through each turn. For a new task, start a new chat — it's a click, and the old one is saved.

When it gets it wrong

HQ reads your question and decides what to search for. A vague question produces a vague search. "Show me good candidates" gives HQ nothing concrete to filter on; "show me candidates with 5+ years of React who applied to the Frontend role" gives it real criteria. Writing good prompts covers this.

A search that returns nothing usually isn't broken — the filter was just too tight, or the data isn't there. Why a search returns no results applies to HQ's searches too. And HQ summarizes what the tools return; for the full record with every field and document, open the profile. The link is right there in the answer.