AI ITSM
About
Useful autonomy. Visible boundaries.
An AI teammate that gathers the evidence, proposes a specific action, and keeps the person responsible in the loop.
The problem
A ticket is the start of the work.
A support request may arrive as one message. Resolving it can require a policy, an employee record, a previous conversation and a change in another system. Someone has to assemble those pieces before deciding what to do.
The original designs brought context retrieval, reply drafting and human review into a ticket workflow. This new interaction study asks a more specific question: what exactly is the person accepting?
A design decision
Make the action as clear as the answer.
The request stays in place while the adjacent workspace moves through evidence, review and execution. A proposal names one operation and its object: create a replacement request for laptop LT-2048.
Accepting the plan and running it are separate states in this exhibit. Editing the plan clears its acceptance. Ordering hardware, sending a message and publishing knowledge remain separate decisions.
Follow the work.
Retrieve the context, supply a pickup location, review the action, and follow one failure through recovery.
When things go wrong
A failed operation shouldn’t erase useful work.
A simulated permission failure preserves the gathered evidence and accepted scope. The reader can inspect the error, simulate an administrator restoring access, and retry the blocked operation.
The final record is a replacement request, with an unsent reply. It is a working design proposition using fictional data, not a measured operational result. The original project’s release history and outcomes remain under review.