Automating IT onboarding and offboarding with AI agents

Automating onboarding and offboarding means turning the joiner-mover-leaver process into structured, routed, and tracked service requests instead of a pile of loose tickets. AI agents handle the intake, split the request into the right tasks for the right teams, chase the missing information, and guard the deadlines. The result: a new employee who is productive on day one, and no accounts of former employees left active for months.

Every IT manager knows the pattern. HR announces a new hire on Friday afternoon, starting Monday. What follows is a scramble of tickets: an account, a laptop, licenses, group memberships, a phone, access to three applications nobody documented. Five teams each own a piece, nobody owns the whole.

Why joiner-mover-leaver work clogs a service desk

Onboarding tickets are individually simple. That is exactly the problem: because they are simple, nobody redesigns them, and the volume quietly eats the desk.

• One request is really ten. A single "new employee" ticket fans out into tasks for identity, hardware, licensing, facilities, and the application owners. When that fan-out happens by hand, steps get forgotten.
• They are deadline-driven. A password reset can wait an hour. A start date cannot move. Every missed step is visible to the new hire and their manager on day one.
• They repeat endlessly, with small variations. Sales gets a different bundle than engineering. The variations live in the heads of two senior agents, which makes the process fragile. This is the same dynamic we describe in recurring tickets automation.

Movers are worse, because half of the request is removal. And leavers are the most neglected of all, which turns them into a different kind of problem entirely.

Offboarding is not a ticket, it is a security control

An onboarding step forgotten is an annoyed employee. An offboarding step forgotten is an ex-employee with working credentials.

Audits keep finding the same thing: accounts that stayed active for months after someone left, licenses still billed, VPN access never revoked. Not because anyone decided that, but because the leaver ticket was free text, went to one team, and the other four never heard about it.

Treat offboarding as a control, not a request. Fixed checklist, every step logged, every step verified, deadline tied to the last working day. That is exactly the kind of structured, repetitive, deadline-driven work automation is made for.

What AI agents automate today

• Intake and completeness. The request arrives as free text from HR or a manager. The AI extracts who, what role, which department, which start date, and asks for whatever is missing in one clarification round instead of three email loops.
• Fan-out to the right teams. The AI classifies the request against your actual catalog and routes each sub-task to the team that owns it, with the deadline attached. A well-maintained service catalog makes this dramatically more reliable.
• Progress guarding. Deadline-driven requests are where SLA compliance breaks silently. The AI flags the sub-task that is going to miss the start date while there is still time to fix it, instead of reporting it afterwards.
• The leaver checklist. Same mechanism, reversed direction, stricter rules: every revocation step tracked, nothing closed until verified, and a human sign-off on the steps that deserve one.

Notice what is not on the list: the AI does not decide who gets access to what. That policy belongs to you. The AI makes sure the policy actually happens, every time, on time.

Where to start

Do not start by automating approvals or provisioning scripts. Start with intake and routing, in shadow mode, on real joiner and leaver tickets. Within two weeks you will see which steps the AI structures correctly and where your own process is the unclear part. Fix the process, then raise the autonomy.

Most organizations discover the same uncomfortable truth: the blocker was never the tooling. It was that onboarding lived in three people's heads. Writing it down for the AI is the first time it gets written down at all.

Frequently asked questions

Can AI create the accounts and assign the licenses itself?

Technically yes, via integrations with your identity platform, but that is step two. The reliable early win is everything around it: complete intake, correct fan-out, guarded deadlines, and a verifiable checklist. Automate the coordination first, the provisioning second.

How is this different from a workflow in our ITSM tool?

A workflow executes fixed steps when someone fills the form correctly. The AI handles the messy reality in front of the workflow: free-text requests, missing information, variations per department, and the chasing when a sub-task stalls. The workflow stays; the AI feeds it clean, complete input.

What is the security benefit for offboarding?

Every revocation becomes a tracked step with a deadline and an audit trail, instead of a line in a free-text ticket. Missed steps are flagged before the leave date, and nothing is closed as done until it is verifiably done. That closes the gap audits keep finding: active accounts of people who left.