Before you post that job description, score the role against these 6 dimensions. The result tells you whether you need a person, a hybrid setup, or an AI workflow.
Are they building, selling, deciding — or mostly tracking and reporting on what others do?
Are the outputs genuinely one-off, or recurring reports, status updates, or approval flows?
Is everything they need already in your tools (logs, CRM, tickets), or locked in conversations?
Are they resolving genuinely messy trade-offs, or applying clear thresholds to structured data?
If AI gets this wrong, is it catastrophic (legal, safety) or cheap to catch and fix?
Can a human review what the AI produces quickly and spot mistakes?
This rubric is adapted from a framework circulating among engineering leaders for evaluating roles against AI automation potential. Read my full take in Stop Hiring Measurers. Build Agents Instead.