AI can replace the expertise needed to perform a task while leaving us dependent on that expertise to make good judgements about the result.

The paradox is that people have traditionally developed that judgement by doing the work AI is now performing for them.

A kind of competence dissonance can emerge when the ability to obtain a result outpaces the ability to judge its outcomes.

However, someone will still need to determine and be responsible for whether the work addresses the right problem, which trade-offs are acceptable, how it fits the wider strategy, and whether it is worth pursuing at all.

PACED is an experiential and reflective framework for developing human judgement under increasing AI delegation.

Develop the underlying knowledge, skills and judgement.

This builds the foundations needed to understand a situation, form an independent view and recognise what a good result looks like.

Examine context, evidence and operational constraints.

This involves defining intent, identifying critical trade-offs, and establishing the boundaries for autonomous execution.

Adjust and refine the approach.

This is an iterative stage of tuning parameters, guardrails, and decision criteria to align agent behaviour with the intended goal.

Assess autonomous execution in practice.

This means auditing agent actions, real-world consequences, and edge-case failures to determine whether the approach is working as intended.

Entrust suitable elements to AI within clear limits.

This involves deciding what can safely be handed off, when humans must intervene, and where accountability remains.

As more trust is placed in agentic routines, judgement and oversight may move further up the stack. Human agency will still be needed to keep outcomes useful, appropriate and aligned with wider goals.