Complaint is a reasonable signal
Analytical approaches to choosing automation are useful and slow. A quicker starting point: what does everyone complain about? Tasks people hate are usually hated for good reasons — repetitive, mechanical, and producing nothing anyone values.
That description is also the profile of a good automation candidate, which is why the instinct is worth taking seriously.
Check it against three numbers
- How often? Daily or more is strong; weekly is marginal; monthly rarely pays.
- How long each time? Timed for a week, not estimated.
- What does an error cost? If mistakes are expensive, the case improves substantially.
If the numbers do not support it, the complaint is about tedium rather than cost. That is still worth addressing — but through process change rather than a build.
The usual suspects
- Copying data between two systems
- Chasing people for information
- Assembling the same report from the same sources
- Formatting documents
- Reconciling two lists that should match
- Answering the same question repeatedly
Every one of these is high-frequency, low-judgement work that a computer does better. If your most-complained-about task is on this list, you have your first project.
The morale return is real
Automation business cases are written in hours and pounds. The return that clients mention a year later is usually different: the team stopped dreading Mondays, or the person doing the worst job got to do something better.
That is difficult to put in a spreadsheet and it affects retention, which is expensive when it fails.
Ask the people who do it
Whoever does the task knows where it goes wrong, which cases are awkward, and what the workarounds are. They are also the ones who will use whatever replaces it.
Involving them early converts potential resistance into ownership, and it produces a far better specification than a manager's description of the same process.
Keep the first one small
One task, one path, four to eight weeks. Prove it, measure it, and let the result fund the next. A first automation that visibly removes a hated job is the best argument for a second.