Your HR & payroll, by conversation
Ask a question, make a request — Copilot answers and gets it done, inside the permissions you already have.
From applying leave to running payroll, Copilot does the work — not just chats — and a human approves anything that matters. Grounded in your real data and your actual policies, so it never makes up an answer.
Answers instantly — and does the work
Employees ping HR for payslips. Managers chase approvals. Admins hunt for missing inputs. The data already exists in Kredily — Copilot closes the gap: ask in plain English, and it answers or acts. One assistant across HR, payroll, attendance, leave and compliance, for every seat.
Answers, instantly
“How many casual leave days do I have left?” “Why was tax deducted this month?” — attendance history, balances, payslips, salary structure, payroll-run status and company policy, read from your real data.
Does the work, not just chat
Apply leave, regularise attendance, claim comp-off / OT, submit expenses, update profile details (address, bank, emergency contacts), onboard employees and manage status changes — real actions, not suggestions.
Run Payroll, guided
Check readiness, surface blockers, calculate, review failures, approve, disburse. Every irreversible step pauses for a human — the full pipeline is below.
Compliance Cop (add-on)
An advisory audit for PF, ESI, PT & TDS — a scorecard with violations, warnings, affected employees and recommendations. It flags issues; it never files returns.
Works where you do
A chat widget inside Kredily, the same Copilot in the employee app, or connected to Claude Desktop and ChatGPT through a secure API (MCP).
Built on 110+ skills
Skills across payroll, attendance, leave, employee records, onboarding and compliance — so a plain-English request maps to the exact right action, every time.
“Run June payroll” — six checkpoints, zero surprises
Copilot walks the run end to end. It never skips a gate, and the irreversible steps wait for you.
Readiness check
Copilot verifies inputs before anything computes — attendance approved, leave resolved, revisions in place.
Blockers surfaced
“3 employees have unapproved attendance” — missing inputs are named, with the affected employees, not buried in a log.
Calculate
Once inputs are clean, Copilot triggers the calculation and reports back when the register is ready.
Review failures
Any employee whose computation failed is listed with the reason — fix and recalculate just those.
You approve
Approval is explicitly human. Approval tokens are single-use and expire — Copilot cannot approve on your behalf.
Disburse
Disbursement pauses for sign-off too. After your go-ahead, salaries pay out through Payout and every step is logged.
Every skill carries a risk grade
Five safety tiers under the hood collapse into three behaviours you’ll notice: instant, confirmed, and signed-off.
Reads — instant
- Leave balances, payslips, attendance history, run status
- Policy questions answered from your actual policies
- Always scoped to what your role can already see
Writes — confirmed
- Apply leave, regularise a punch, comp-off / OT and expense claims
- Profile updates, onboarding and status changes
- Copilot prepares the change; you confirm before it commits
Critical — signed off
- Payroll approval, disbursement and compliance actions
- Explicit human sign-off; approvals single-use and expiring
- Every action written to the audit trail
AI that acts — without acting up
The #1 fear with AI that takes actions is that it does something wrong. Copilot is built so it can’t.
Never more than you can
Copilot inherits the exact role-based permissions of the logged-in user. An employee’s Copilot can’t approve payroll — and no phrasing changes that.
Humans approve risk
Payroll runs, disbursements and compliance actions always pause for explicit human approval. Approvals are single-use and expire.
Read vs. write is graded
Routine reads happen instantly; sensitive changes need confirmation; critical actions need sign-off — five safety tiers under the hood.
Full audit trail
Every action Copilot takes is logged — what changed, by whom and when.
Your data stays yours
Strict per-company isolation, processed in-region (India-resident AI by default). Keys are encrypted at rest.
Tunable autonomy
Each company chooses how much Copilot may do on its own — from “observe only” up to “autonomous”, with critical actions still gated.
An advisory auditor on retainer
An optional, separately priced add-on that audits your statutory posture — and tells you exactly what to fix.
Four statutes audited
PF, ESI, Professional Tax and TDS are checked against your actual payroll data — not a generic checklist.
A scorecard, not a shrug
Violations and warnings, each with the affected employees named and a concrete recommendation to resolve it.
Before the deadline
Issues surface while there’s still time to fix them — before the penalty, not in the audit that follows it.
Advisory, by design
Compliance Cop flags and recommends. It never files or submits returns — statutory filings stay with your team and your CA.
Grounded, routed, and gated
Reads your real data
Copilot reads live data and your actual company policies — auto-generated into its knowledge base — so answers are grounded, never guessed.
Routes to the right model
Multi-model (Claude, Gemini, OpenAI) and cost-smart: simple questions use a fast, economical model; high-stakes payroll and compliance reasoning escalates to the most capable one. Deterministic checks skip the AI entirely.
Acts within permissions
It performs the action inside your role’s permissions — and pauses for a human to approve anything irreversible, logging every step.
Three doors, one Copilot
The same skills, the same permissions, the same audit trail — wherever you ask from.
Inside Kredily
A chat widget on every authenticated screen that already knows which module you’re in and who you are.
In the employee app
The same Copilot rides along in the Kredily employee app — balances, payslip questions and leave requests from a phone.
In your AI tools, via MCP
Connect Claude Desktop, ChatGPT and other MCP-capable clients through a secure key. Same role, same permissions — a different door into the same guarded house.
One Copilot, every seat
Employees
“Apply 2 days sick leave from Monday.” Regularise a missed punch, claim comp-off / OT, submit an expense, check a payslip — just by asking.
Managers
“Who’s on leave next week?” Approve or reject leave and comp-off requests, see team status, get instant answers about reportees.
HR & Admin
“Onboard Priya as a Software Engineer from the 1st.” Update records, manage exits and get instant answers across the whole organisation.
Finance & Payroll
“Are we ready to run June payroll?” A guided, approval-gated run that surfaces input blockers and failed employees before disbursing.
Copilot has a sibling: Workforce Intelligence
Two AI products, one principle — the AI never makes up numbers.
Copilot converses
Questions and actions, one person at a time — grounded in the live data that person is allowed to see.
WI watches the whole
Headcount, attrition, payroll cost and absenteeism as computed metrics, with early-warning insights that connect signals across modules.
Same data, same honesty
Every number comes from the metric engine; the AI only explains it. Explore the leadership cockpit in Workforce Intelligence.
Copilot questions, answered
Can Copilot do something I’m not allowed to do?
Will it run payroll on its own?
What actions can Copilot actually take?
Can it make up an answer?
Is my company’s data isolated?
Is there an audit trail?
Where do I use it?
Which AI models power it?
Does it handle compliance?
Explore the rest of the platform
Ask. Approve. Done.
See Copilot answer and act on your own HR & payroll data — safely, within the permissions you already have.