The four tests
High repeatability, low judgment, steady volume and recoverable errors point to a system. The inverse points to a person. Mixed answers usually mean the task should be split, with the clerical half automated and the judgment half kept.
- Repeatability — does it happen the same way most of the time?
- Judgment — does a good outcome require reading a situation?
- Volume stability — is it constant, or a seasonal spike?
- Risk if wrong — what is the cost of a mistake nobody catches?
Split the role, not the person
Take an office coordinator. Sending appointment confirmations, logging outcomes and generating invoices are rule-based. Calming an upset customer, negotiating a reschedule with a difficult homeowner and spotting that an estimate looks wrong are not. The right answer is rarely 'automate the coordinator' — it is 'stop making the coordinator do clerical work'.
The costs people forget
- Systems: configuration, maintenance, exception handling, and change management.
- Hires: recruiting, onboarding, management time, turnover and coverage.
- Both: the cost of the work being done badly during the transition.
When to hire even though you could automate
- You need capacity this month and the system will take a quarter.
- The process is still changing weekly and would be automated wrong.
- The work is customer-facing and the relationship is the differentiator.
- Regulation or licensing requires a named, accountable person.