Challenges

The prompt isn't the hard part

Wording helps. Goals, constraints, and verification decide whether the work holds.

The trap

People chase magic phrases because phrases feel controllable. The harder work is knowing what “done” means, what evidence you need, and who owns the outcome.

A mediocre prompt with a sharp acceptance test beats a poetic prompt with none.

What this means at work

If you spend an hour polishing adjectives and zero minutes naming the check, you optimized the wrong layer. Prompts steer. Acceptance criteria decide.

Practice (10 minutes)

  1. Take one AI-assisted task you will do today.
  2. Write three bullets: what good looks like.
  3. Write two bullets: what bad looks like (hype, invented policy, missing owner).
  4. Put those bullets in the prompt. Generate once. Score only against the bullets.

Reflection

Where did you last blame the prompt when the real gap was an unstated finish line?

See also