- The 6-step prompting framework: Who am I / Task / Audience / Output / Avoid / Upload
- The two most underused steps: Avoid (the escape hatch from generic AI voice) and Upload (giving AI your actual source material rather than asking it to generate from nothing)
- Worked examples for each step, applied to real work tasks across functions
- The doing-to-directing reframe: you're not asking AI to do a task — you're directing it the way you'd brief a highly capable but completely uninformed colleague
- Why iteration is the skill — first prompts almost never produce final outputs, and that's the point
- How to build a reusable prompt for tasks you do repeatedly, instead of starting from scratch each time
Pick one task you repeat every week. Write a reusable prompt using the 6-step framework — fill in each step deliberately.
Run it twice this week, on the same kind of task. Notice what changes between run one and run two — and what stayed the same.
Discuss the prompts we built this week — what worked, what needed refinement, and which ones might be useful across the team.
Prompts that started rough and got better with iteration. Moments where adding the audience, or what to avoid, or attaching source material made the difference. Prompts that didn't work even after refining — those are usually missing context the AI needed.
- How long should a good prompt be?
- Long enough to give context, short enough to run in under two minutes. If you're writing an essay, you've overcomplicated it. The 6 steps are the discipline.
- Do I have to use this framework every time?
- No. Once the thinking becomes instinct, you stop needing the framework. It's training wheels, not a permanent requirement.
Send us the 2–3 best prompts your team built this week — with the task context, not just the prompt text. The most useful ones get curated into a shared library for the function.