01● Keep human
Decide the exit terms: last day, notice, pay and any restrictions
Before any task is triggered, someone accountable has to settle the facts of the exit: the last working day, whether notice is worked or paid in lieu, what is owed in final pay and accrued leave, and whether any non-compete, confidentiality or non-solicitation terms apply. These decisions carry legal weight and differ by state, contract and reason for leaving, so they stay with HR or a manager with authority. AI can summarize a contract clause or draft a plain-language note of the options, but it should not decide the terms or promise anything to the employee.
⛨Applies to every departure. Voluntary resignations are routine; dismissals, redundancies and settlement agreements need legal review before anything is communicated.
02● Standard automation
Create the exit checklist and assign every task to an owner
Once the last day is confirmed, the checklist should appear on its own: IT access, equipment return, payroll, benefits, building access, handover, client introductions and the exit conversation. This is a fixed template with branches (remote or office, manager or individual contributor, contractor or employee), which is exactly what rules-based workflow in your HRIS or ticketing tool does well. Each task gets an owner and a due date relative to the last day. AI adds little here and can silently drop a step, so keep it deterministic and auditable.
⛨Works when the template is maintained and the HR system or ticketing tool is the single source of the leaver's data. Review the template at least twice a year.
03● Standard automation
Remove system access and recover equipment on the last day
Disabling accounts, removing single sign-on, revoking licenses, forwarding email, transferring file ownership and collecting laptops, phones and badges should run from the checklist with a timestamp for each action. Identity tools can do most of this automatically at a set time. A person still confirms that physical items came back and that shared accounts, API keys and vendor portals which are not tied to single sign-on were also rotated. Missed access is the most expensive offboarding failure, so the log matters more than the speed.
⛨Needs an up-to-date inventory of the systems and devices each person uses. Sensitive roles may require access to be cut at the moment of notice, not on the last day.
04● AI candidate
Draft the handover note and capture what the leaver knows
AI is useful for turning the leaver's open tickets, recent emails, shared documents and calendar into a first-draft handover note: active projects, open commitments, key contacts, recurring deadlines and where files live. The leaver and their manager then correct it, which is much faster than writing from scratch. The draft should only draw on sources the leaver and the company are allowed to use for this purpose, and the manager must read it before it is shared with the successor.
⛨Works best when work happens in trackable tools. Where the knowledge sits in someone's head or in private notes, an AI draft will be thin and a conversation is needed.
05● Human review
Hold the exit conversation
The exit interview is a human conversation. A leaver will say more to a person they trust than to a form, and the manager or HR contact needs to hear tone, hesitation and the things that are not said. AI can prepare a short list of questions and, with consent, produce a transcript or notes, but it should not run the conversation. Keep it voluntary, separate from the final pay discussion and clear about who sees the notes.
⛨Always for voluntary leavers who agree to it. Skip or adapt it for dismissals and disputes, where legal advice should come first.
06● Human review
Handle final pay, leave balance, benefits and references
Final pay rules vary by state and country, and can include accrued leave, bonuses, expenses and benefit continuation. Payroll software calculates the numbers, but a person checks them against the contract and the applicable rules and signs them off. References and any statement about performance should be written or approved by the manager, because a careless line is a legal risk. AI can draft a neutral confirmation of employment dates and title for a person to review.
⛨Required for every leaver. The deadline for final pay depends on the jurisdiction, so check the rules for the state or country involved.
07● AI candidate
Analyse exit feedback and attrition patterns
Across many departures, AI can group free-text exit feedback into themes such as pay, management, workload, growth and flexibility, and flag a team or manager that keeps appearing. That is pattern-finding on text, which AI does quickly. The conclusions are hypotheses, not findings: a person decides whether a pattern is real, what to do about it and how to protect the anonymity of small groups.
⛨Only useful with enough departures to show a pattern, and with exit data stored consistently. For fewer than about ten leavers a year, read them yourself.