A practical scoring method for ranking automation candidates by value, feasibility, and the likelihood anyone actually adopts the result.
The wrong way to choose what to automate is to pick the process that annoys the loudest person. The right way is to score candidates on four dimensions and be honest about the inputs.
How often does this task occur, and how long does one occurrence take? A weekly task taking two hours is 100 hours a year. A daily task taking twenty minutes is 87. Frequency matters more than duration, which is why nobody notices the small ones.
Automation thrives on predictable structure. If every instance is genuinely different — bespoke client work, novel legal questions — automation will handle a fraction and the exception queue will eat the gain. Score variability honestly; this is where most business cases quietly fail.
A mis-routed support ticket is recoverable. A mis-stated invoice or a wrong clinical record is not. High-error-cost processes are still automatable, but the design must include human confirmation at the consequential step. That changes the value calculation.
The best automation in a spreadsheet nobody opens is worth zero. Ask who has to change their behavior for this to work, and whether they have a reason to. Workflows that remove visible pain get adopted. Workflows that add a step for someone else's benefit get bypassed.
For each candidate, rate 1–5: frequency, time cost, predictability, error cost (inverted), and adoption ease. Multiply frequency × time cost × predictability for a raw value, then adjust for error cost and adoption. It is not a precise model — it's a way to force the conversation onto the dimensions that decide outcomes.
Do not start with the most complex process, even if it's the most expensive. Start where you can ship, measure, and be believed.
Keep going
A strategy call starts with the workflow you're already thinking about.
