An HR helpdesk assistant grounded in your actual policies

Grounded answers, not a chatbot guessing

The assistant answers only from your real sources: your HR handbook, policy documents, and approved FAQs, retrieved at question time so the response reflects what your policies actually say. This is retrieval-grounded generation, not a model recalling something vague from training. When an employee asks about notice periods or expense limits, the answer comes from your current document, and the assistant can point to where it found it so the employee can trust and verify it.

Just as important is what it does when it does not know. Rather than improvising, the assistant is built to say when a question falls outside its sources and hand off to a human. That honesty is the whole point: an HR assistant earns trust by being reliably right on the routine questions and reliably cautious on everything else, never by confidently making something up.

• Answers drawn from your handbook, policies, and approved FAQs
• Retrieval-grounded at question time, reflecting current policy
• Cites where an answer came from so employees can verify it
• Says when it does not know instead of improvising
• Kept up to date as your policies change

Knowing what to escalate, and to whom

Plenty of HR questions should never be handled by a machine, and the assistant is designed to recognise them and route them to a person. Anything sensitive, a grievance, a health issue, a pay dispute, a conflict with a manager, goes to the right member of your HR team rather than getting a canned reply. The same applies to anything the assistant is not confident it can answer from your sources.

This is the human-in-the-loop line applied to employee support: the assistant handles the high-volume, low-stakes questions so your HR team is not buried in them, and it deliberately steps back the moment a situation needs empathy, discretion, or judgement. Employees get fast answers to the routine and a real person for the things that matter, which is a better experience than either a purely manual queue or an all-automated bot pretending to be helpful.

• Sensitive topics like grievances, health, and pay always go to a person
• Low-confidence questions escalate rather than getting a guessed answer
• Routes each escalation to the right HR owner with context attached
• Frees HR from repetitive questions without abandoning the hard ones

Private by design

An HR helpdesk handles some of the most sensitive data in the company, so privacy is not optional. We build it to run in your own environment, using models and infrastructure that keep employee questions and personal data under your control rather than leaking them to a public service. Access is scoped so the assistant can see the policies it needs and only the personal data appropriate to the question and the person asking.

Where a question needs personal specifics, such as an individual's remaining leave, the assistant works against your systems with proper access controls and logging, so it never becomes a way to pull data someone should not see. Under the GDPR that discipline is not just good practice, it is required, and we design for it from the start rather than bolting it on.

Wired to your systems, owned by you

The assistant meets employees where they already are, in Slack, Teams, or an internal portal, so there is no new tool to adopt. Behind that, it connects to your knowledge sources and, where useful and permitted, your HR systems, so it can answer both general policy questions and specific ones like a leave balance. When it escalates, the ticket lands in your normal HR queue with the conversation attached.

As always, you own it. The knowledge base, the escalation rules, the integrations, all documented and running in your accounts, so your HR team can update a policy, change what escalates, or adjust the assistant's scope without depending on us. It is your HR knowledge made instantly accessible, not a black box you have to trust blindly.