Offboarding must include the context your AI tools can reach
Include AI connectors, shared projects and stored context in employee offboarding, using an explicit access inventory and verification.
Removing someone from the team directory may not describe every place they can still access company information. AI workflows can involve shared projects, external tools, connectors and credentials with their own ownership.
Include those access paths in the existing offboarding process.
Inventory the actual connections
Record which approved AI tools the person used, how they signed in and which company sources were connected. Identify shared workspaces, individually owned automations and credentials used by integrations.
Use the vendor's current documentation to determine how access revocation works in each tool. Do not assume that disconnecting a source deletes previously imported copies or that removing one account disables a separately owned integration.
Transfer work before removing its owner
Assign a new owner to shared workflows, knowledge collections and pending reviews. Preserve company records through the approved process while keeping personal material and access restrictions separate.
Check whether a service depends on a departing employee's identity. Where appropriate, move it to a supported company-managed identity rather than sharing the person's login.
Verify the outcome
After access changes, verify them using administrative records and approved test accounts. Check the source system and the connected tool. A completed checklist entry is not evidence that a connector stopped working.
Record unresolved vendor limitations and route them to the owner responsible for the relevant information. Avoid promising deletion or revocation behaviour that has not been verified.
Add AI access to the same joiner, role-change and leaver process used for the rest of the company. Treating it as a routine part of account management makes it easier to keep complete than relying on a special cleanup after someone leaves.