Ask in your own words. Klokee does the paperwork.
An employee types "I need Friday off". Klokee works out the leave type, checks the balance, drafts the application and shows it. If nobody confirms it, nothing happens — the agent has no authority to write on its own.
It Drafts, You Decide
Leave applications, punch corrections, expense claims and loan requests are prepared as drafts. Confirm is a human action, always. This is the single rule the whole agent is built around.
It Knows Your Policy
Klokee reads the actual leave policy, balance and holiday calendar for that employee — not a generic answer about what leave usually works like.
Plain Language In
No form to find, no field to learn. The employee describes what they need the way they would say it to a person.
Wrong Drafts Cost Nothing
A draft that misread the request is discarded in one tap. That is the point of separating drafting from committing.
An assistant that files, not one that answers
Most HR chatbots retrieve. Ask about your leave balance and you are told the number, which is useful once. Klokee is built to do the next thing: work out what you are asking for, prepare the actual application, and put it in front of you complete.
"I need Friday off" becomes a leave application with the type chosen against your policy, the balance checked, the dates set, and the approver identified. You read it and confirm, or you change it.
The boundary is absolute
Klokee drafts. It does not write. There is no configuration in which it approves a request, files a claim, or alters a record on its own — the confirm step is the product, not a safety setting bolted on afterwards.
This is what makes it usable by people who do not trust software with their salary. The worst outcome of a misunderstanding is a draft you decline, which costs one tap.
What it can prepare
Leave applications, punch corrections, expense claims and loan requests, in plain language, in the employee's own words rather than in the vocabulary of the form. That last point matters on a shop floor: the person who cannot find "regularisation" in a menu can certainly type that they forgot to punch out on Tuesday.
From a sentence to a filed request
They type
Plain language, any of the supported languages, from the phone they already have.
Klokee works it out
Intent, leave type, dates, balance, policy, approver — checked against the actual configuration for that person.
A draft appears
The complete request, shown in full. Wrong ones are edited or discarded at no cost.
They confirm
Only now does it become a real request and enter the approval chain.
Ask Klokee FAQs
Can Klokee approve or change anything by itself?
No, structurally. Klokee drafts — a leave request, a punch correction, an expense claim, a loan request — and writes only when a human presses Confirm. It cannot approve, and approval stays with whoever your policy names, through the same chains with the same cover and escalation. An assistant that acts unsupervised inside payroll is a liability with a friendly name, and we declined to build one.
What if it drafts something wrong?
You discard it, and nothing has happened — no record written, no approval consumed, no trace beyond the seconds it took to read. That is the practical meaning of the confirmation step: the cost of a wrong draft is zero, which is what makes it safe to let the assistant attempt things. The audit trail records what was confirmed and by whom, so responsibility stays with people, where a dispute needs it to be.
Does it need training on our policies?
No. Klokee reads the policies you have already configured in Klok — leave rules, balances, holiday calendars, shift patterns — so its drafts are computed against your actual rules, not a generic model of an average company. There is no separate knowledge base to feed and no drift between what the assistant believes and what payroll will do, because both read the same configuration.
What can employees actually ask it?
The things they currently ask HR: what is my balance, apply for two days' leave, fix yesterday's missing punch, file this receipt. Each becomes a draft against their own record, in their own language, on their own phone — and the routine traffic that eats an HR team's mornings becomes confirmations instead of conversations. What it will not do is answer for another employee's data; scope follows the person asking.
Is this on our bill if nobody uses it?
Modules are billed only when switched on, so an assistant nobody wants costs nothing. In practice Klokee earns its place with the workforce that will not open a web form but will send a message — which for most Indian SMBs is most of the workforce. Try it with one team before deciding for everyone; the switch is per company and reversible.
Does Klokee see data it should not?
No — it computes against the asking employee's own record and your configured policies, through the same permission boundary as every screen. It cannot be talked into revealing another person's salary or balance, because scope is enforced where the data is read, not in the conversation. An assistant is only safe to deploy if it is exactly as constrained as the person using it.
See it on your own data
Fourteen days, no card. Load a handful of employees and judge it against how you work today.