The Administrative Automation Problem
Which repetitive administrative workflows could be safely automated while preserving necessary human oversight?
The problem
Where students run into it today.
The same form, the same signature, the same six steps.
Why it matters
Repetition is expensive in a way nobody measures, because the cost is spread across everyone a little at a time.
Automation without oversight is worse than the manual process. The design question is where the human stays, not how to remove them.
It forces a genuinely useful discipline: mapping a process before writing a line of code.
Possible directions
Several credible approaches, no obvious winner. Choosing between them is part of the work.
Process mapping first
Document an existing workflow end to end, including the exceptions. Most of the value is created here, before any automation exists.
Status transparency
Many enquiries exist only because a request is a black box. Showing where something is removes the need to ask.
Human-in-the-loop by design
Automate assembly and routing; keep judgement, approval and exceptions with a person, explicitly and visibly.
Technologies & difficulty
Real users, real failure modes. Reliability, privacy, access control, monitoring and the discipline to keep it running.
Difficulty describes the work, not the person. A level 04 problem is not closed to a first-year — it just needs a team and a longer runway.
Potential project
Campus Automation Engine
A workflow layer that handles the routing and status of repetitive requests while keeping approval and exception handling with a person.
No team has taken this on yet.
Skills needed
Open roles & possible team
- Backend
Workflow engine, audit trail, integrations
- Security
Access control and data handling review
- Technical Writing
Process documentation — the real deliverable
Research questions
The things nobody in the group can currently answer.
Which steps in a workflow carry judgement, and which are purely mechanical?
What has to be logged for an automated decision to remain reviewable months later?
How do you fail safely when an automated step is wrong?
Next steps
What the first two weeks would sensibly look like.
- 1
Pick one workflow. Document every step, every exception and every person who touches it.
- 2
Publish the map. If nothing gets automated, the map is still worth having.
- 3
Automate only the steps that nobody disagrees about.