What to Automate First: A Simple Hours-Versus-Effort Test

Most businesses have a long list of tasks that feel automatable. Automating all of them is a mistake. The goal is to find the few where the payoff is clear. This is the test we use when scoping business process automation work.

Score each task on four questions

1. How often does it happen? A weekly task costs 52 repetitions a year. A daily one costs about 250.
2. How long does it take each time? Include the interruptions, not only the keystrokes.
3. How rule-based is it? If a new hire could do it from a written checklist, it is a good candidate. If it needs judgment on every case, it is not.
4. What does a mistake cost? Re-keyed invoices and order data are error-prone, and the errors are expensive to find later.

Multiply frequency by time to get hours per year. Then ask what it would take to build and maintain a fix. The best first project has high hours, clear rules, and a simple build.

Good first candidates

• Copying data between two systems that both have an API
• Weekly or monthly reports assembled from the same sources
• Pulling fields out of emails, forms, or PDFs into a system of record
• Routine notifications and hand-offs that currently depend on someone remembering

Poor first candidates

• Rare tasks, however annoying
• Processes that are still changing every month
• Work that depends on judgment calls that have never been written down

Sometimes the answer is to simplify

A process with seven approval steps is not improved by automating all seven. Mapping the workflow first often shows steps that can be removed, which is cheaper than building around them.

If you have a candidate and want a practical read on whether it is worth building, tell us what's broken.