AI Ready
Prompts card, MethodKit for AI Readiness
Card 35 of 48 · MethodKit for AI Readiness
  • ThemeLevel Up
  • CardCard 35 of 48
  • Questions5 to explore
Level Up

Prompts

How you formulate instructions so AI gets it right

A prompt is an instruction, and the quality of what AI produces is mostly a function of how clearly you told it what you actually wanted.

A vague prompt produces a generic answer. A precise prompt that names the task, the format, the audience, and the constraints produces something you can actually use. The gap between those two outcomes is often wider than people expect, and it is almost never the model's fault.

Good prompts are not magic words or tricks. They are the same things that make any briefing work: context about the situation, a clear description of the output you want, any constraints on format or tone, and enough background that the reader could do the task without guessing. The difference is that AI will not ask follow-up questions to fill in the gaps. It will just fill them in, often incorrectly.

Prompting is a skill that gets better with practice, and it compounds. Teams that build a library of prompts that work can reuse and refine them. Teams that write fresh prompts every time are always starting from scratch.

Make it visibleTake the last task you asked AI to help with and rewrite the prompt as if you were briefing a capable but uninformed colleague: include the task, the audience, the format, and two constraints. Run it again and compare the output.

Why AI needs this

Each part of your work matters to AI in a specific way. Some of it is context a tool needs before it can help, some of it is work a tool can take on, and some of it is judgment that should stay with you.

Task framing

AI needs to know what kind of task this is. 'Write something about the product' and 'Write a 150-word product description for a retail buyer unfamiliar with the category' produce very different results.

Role and tone

Telling AI to write as a particular kind of person or in a specific register anchors the output. Without it, the default is generic and slightly corporate.

Output format

AI will choose a format if you do not specify one. Specifying the format (bullet points, numbered steps, a table, a paragraph) saves a re-draft.

Constraints and edge cases

What the output should not do is often as important as what it should. Telling AI what to avoid stops it from doing the plausible-but-wrong thing.

Questions to explore

Use these on your own or in a group. There are no right answers, only better conversations.

  1. When AI gives you a bad result, do you usually know why the prompt produced it?

  2. Which prompts does your team use repeatedly that are not yet written down anywhere?

  3. What context do you typically have to re-explain every time because it is not in the prompt?

  4. Who on the team gets the best results from AI, and what are they doing differently?

  5. If you shared a prompt with a colleague, would they get the same kind of output you expect?

Readiness traps

  • Blaming the model when the brief is the problem. Most poor AI output is a prompt problem, not a capability problem. Rewrite the prompt before you conclude the tool cannot do it.
  • Keeping prompts that work in one person's head or browser history. A prompt that is not saved and shared is one bad day away from being lost.
  • Treating a prompt as done when it works once. Good prompts are revised over time as you learn what produces consistent results across different uses.