HR first line
The question about the health plan, answered before Monday.
- Collective agreement and rules
- Sourced answer
- Escalation to a human when needed
It reads your collective agreement, your internal rules and your procedures. It returns a first-line answer, and the escalation when one is called for. It handles neither disciplinary nor medical matters.
The same question, every Monday.
The health plan, leave, sick pay, travel costs. The answers exist, written down, in an agreement and a set of rules nobody opens. They end up going through one person, on top of their own job.
This is not an availability problem. It is a document nobody can find.
3 guardrails, not 3 promises.
Each one proves itself in use, not in a brochure.
It answers from your own texts
Collective agreement, company agreements, internal rules, staff notices. The answer cites the article and the document, so the employee can read it themselves.
It escalates sensitive matters
Disciplinary, health, personal circumstances, conflict: it does not handle them. It points to the right person, without keeping the detail of the request.
It does not improvise a rule
What your documents do not cover, it does not infer. It flags the gap, which lets you close it once and for all.
The first line absorbed, the rest redirected.
The employee asks their question the way they usually would. The assistant finds the applicable article in your documents, answers citing its source, and stops dead as soon as the subject needs a human. You get back a list of the questions asked, which shows where your texts are vague.
The rest, once trust is settled.
What never leaves your walls.
If your AI provider cannot put these four lines in writing, ask them why.
Watch it work on your document.
Twenty minutes, on your own data. Bring what this assistant will have to read, and you watch live what it does with it, and where it stops.