The expensive part of HR is not designing the policy. It is answering the same question for the fortieth time.
How much leave do I have left. How much overtime this month. When does this compensatory day expire. Who is my leave request sitting with. Each takes two minutes, and each has an answer that is already in the system.
What it costs now
| Input | Example |
|---|---|
| HR queries per day | 25 |
| Of those, answerable from the system | 18 |
| Minutes each: read, look up, reply | 2 |
| Per day on already-answerable questions | ≈ 36 m |
| Per month, 22 working days | ≈ 13 h |
Thirteen hours is a week and a half of one person's time — in dozens of interruptions rather than a block.
Once it is connected
| Now | After | |
|---|---|---|
| Queries reaching a person | 25 | ≈ 7 |
| Time per day | 50 m | ≈ 14 m |
| Recovered | ≈ 36 m/day |
The already-answerable layer drops close to 100%, and total query handling falls by around two thirds.
The denominator is the 18 queries that already had an answer. The remaining seven still need a person — and those are usually the ones that need judgement: how an unusual leave type is calculated, whether to make an exception, how to handle a payroll dispute. Giving the time back to those is the point.
What it actually does
- Answers about your own record: leave balance, overtime hours, expiry dates, payslip detail.
- Runs the process: leave, overtime and correction requests submitted in plain language instead of hunting for form fields.
- Reports approval status: who it is with, and for how long.
- Summarises for HR: who is approaching the overtime limit, which balances expire soon, which requests have been unsigned for more than three days.
Permissions are the centre of this scenario
HR data is the most sensitive category in most companies, so the design has one rule:
An answer only appears within what the person asking may already see.
Employees see their own, not the colleague's; a manager sees their own team; HR sees whatever their existing role allows. Permissions are inherited from the systems you already run rather than redeclared for the AI — so nothing depends on anyone remembering to tag a record correctly.
What it will not do on its own
Finalising payroll, transferring money, deleting records, issuing letters — these committing actions always wait for a person. The AI retrieves and organises; the decision stays where it was.
This is a scenario on QSyn
This is not another HR system. It is what QSyn does once it is connected to the attendance and payroll data you already keep — the same knowledge and the same permission model that also serve support, service and scheduled checks.
One internal system we operate runs exactly this way: attendance, payroll, leave, overtime and the staff directory, with an embedded assistant operated in natural language and strictly bounded by each user's own permissions.
Common questions
Do we have to replace our HR system?
No. It sits on top of what you already run, reading and writing through APIs or a suitable integration. Existing processes stay as they are.
Could an employee see someone else's salary?
No. Answers are limited to what the person asking may already see, and that limit comes from your existing permission configuration rather than a separate list that somebody has to maintain.
Is our data used to train models?
No. The underlying models are used under enterprise agreements that exclude customer input from training. The platform can also be deployed inside your own environment so that nothing leaves the company.
What if staff will not use it?
It lives in the messaging tool they already use, so there is no new software to learn. We walk the team through it in person when it goes live.
