Calibration as an Ongoing Practice

Your calibration is not fixed. It should change as you learn more about the executor's capabilities and your own standards. If you consistently find that you're accepting work that turns out to be wrong, you're over-trusting. Increase your scrutiny. If you consistently find yourself rewriting work that actually met your standard, you're under-trusting. Decrease your scrutiny.

This learning is most visible with AI. Different AI systems have different strengths and failure modes. One might be very good at summarization but poor at novel analysis. Another might be excellent at working with structured data but weak at handling qualitative nuance. As you direct more work to a specific AI system, you learn its profile. Your trust calibration should reflect that learning.

The same principle applies with human colleagues. You learn over time what they're reliable for and what they need more support on. The difference with AI is that the learning often needs to be more systematic and explicit. You can't rely on hunches or intuition built up over years of casual interaction. You have to pay attention, notice patterns, and adjust.