What are you trying to do?
Name the job and its purpose. “Draft a reply to this customer’s delivery question” is clearer than “Help with this.”
A PRACTICAL AI HABIT
Generate.
Ground. Review.
Give clear instructions. Use the right information.
Check the result before you rely on it.
START WITH YOUR WORK
Choose one task to practice.
Pick a real task you can check: an email from meeting notes, a summary of a policy, or a comparison of two proposals. Keep the relevant material ready in an approved workspace.
What would someone need to know to do this well?
If a teammate would need certain information to do the task well, ChatGPT probably needs it too. Think about the goal, audience, background, sources and constraints before asking for a first draft.
Read the key ideas and practical tipsHOW A LANGUAGE MODEL WORKS
A useful response needs more than a question.
A large language model generates responses based on patterns, instructions and context. Better results come from grounding it with the right information and reviewing the output before you use it.
Hallucination /həˌluːsɪˈneɪʃən/ · Noun
A plausible response generated by an AI model that is not properly grounded in the right information.
Use ChatGPT as the workspace, not the source of truth. Start with trusted files, authoritative records or web search when you need current facts, policies, legal material, prices, dates, citations or exact source text. Search means checking information online.
Then ask AI to summarize, explain, compare, organize or draft from the material. Check that each important claim is actually supported.
HABIT 01
Task. Context. Expectation.
What information should ChatGPT use?
Provide the relevant records, sources, examples, audience and constraints. Identify missing information instead of inviting a guess.
What would a good response look like?
Specify format, length, tone and the checks that matter. Ask for source references and unresolved questions where appropriate.
A simple example
Task: Draft an email.
Context: It is for a colleague who missed today’s project meeting and needs the key decisions.
Expectation: Use three bullets, a warm tone, and a clear next step.
Supply the actual meeting notes. The model should not invent the decisions.
HABIT 02
Review the output with CLEAR.
Complete
Does it answer every part of the task and include required details?
Logical
Do the reasoning, calculations and conclusions make sense?
Evidence
Are important claims supported by the right sources, and do those sources actually say that?
Audience
Is the language, detail and tone right for the people who will use it?
Relevant
Does it serve the actual goal without distracting or unsupported material?
Correct what needs work. Ask for missing evidence. Use your judgment before sending, publishing or acting on the result.
RECAP
Build the habit.
Check each box to continue to the prompt builder.
0 of 3 checked
Write a better prompt.
Your notes stay in this page. This builder formats your instructions; it does not send them to an AI service.
After you use the prompt, review the answer with CLEAR: Complete, Logical, Evidence, Audience, Relevant.
Sources and further reading
PUT IT TO WORK
What should we build for you?
Bring one bottleneck. We’ll discuss the sources, the system and the business case.