Automation programmes often begin with a platform decision and then hunt for processes to justify it. Reversing the order produces better results: quantify the work first, then choose the lightest mechanism that can carry it.
The arithmetic is unglamorous — volume, handling time, error and rework rate, cost of delay — but it is what separates automation that compounds from automation that becomes another system to maintain.
Quantify before you design
- Cases per month and their seasonality
- Average and worst-case handling time, including chasing and rework
- Cost of an error, and who absorbs it today
Bound the variability
Automation is cheap where inputs are structured and rules are stable. Where variability is high, the honest design is a narrow automated path plus a well-designed exception queue, not a rule engine that grows forever.
Choose the lightest mechanism
In order of preference: remove the step, integrate the systems, orchestrate with a workflow engine, then apply a model. Each step down that list adds capability and maintenance cost in equal measure.
Track benefit after go-live
Instrument the same measures you used to justify the work and review them a quarter after launch. Programmes that publish their own results earn the mandate for the next tranche.