I get this question constantly from business leaders. Here's my back-of-the-napkin test that works every time:
Look at two dimensions: How accurate can the AI be, and what's the cost if it gets something wrong? A dollar, a customer, or the whole company?
📊 Low accuracy + Low cost = Perfect assistant (ChatGPT for brainstorming, collecting info - but not making decisions)
🤖 High accuracy + Low cost = Pure automation (routing tickets, moving data point A to B - no human touch needed)
👨💼 Low accuracy + High cost = AI provides input only (I'd never let AI hire a senior executive without me meeting them!)
✈️ High accuracy + High cost = Human in the loop (Like airplane autopilot - AI does 80-90% of the work, but pilot has hands on the wheel for final judgments)
These four categories help you internalize exactly how much to automate versus how much AI should just be helping you and your team make better decisions.
The goal isn't to eliminate human judgment - it's to amplify it in the right places.