Before you buy another AI tool, map one day of work.
A practical way to find a useful AI starting point: follow one task, identify the friction, and define a result worth measuring.
You do not need a theatrical prompt to explain a familiar task. Start with the information you would give a capable colleague doing the work for the first time.
Write down what needs to be produced, who it is for, and what the reader should be able to do with it. “Write something engaging” leaves almost every useful decision unresolved. “Draft a short internal update explaining the approved change and the action each team needs to take” provides a real assignment.
Keep the goal separate from the format. The goal might be helping a colleague understand a decision. The format might be three paragraphs and a short action list. A neat format cannot compensate for an unclear purpose.
Provide the approved information needed for the task, subject to your organization’s rules and the system’s approved use. Say which document controls if two sources disagree. If a required fact is missing, instruct the assistant to flag the gap rather than fill it with a plausible guess.
Do not paste sensitive or confidential information into a tool simply because the task would be easier with more context. Check that the particular tool and its settings are suitable for the information and use. The linked Cyber Centre guidance explains why information handling needs attention.
Useful constraints are concrete. Keep the agreed dates unchanged. Do not add a quotation. Distinguish confirmed actions from suggestions. Use the supplied terminology. Mark unanswered questions in a separate section. These instructions describe what would make the draft usable.
A style reference can help, but describe what you value about it: short paragraphs, a direct opening, or a calm tone. Asking for a voice without explaining its features makes the result harder to assess.
Purpose: explain an approved project change to the delivery team. Audience: colleagues who know the project but missed the update. Source: the approved change note. Output: a short explanation, the confirmed actions, and any unanswered questions. Boundaries: do not invent dates, owners, or commitments. Review: the project owner checks the draft against the source before use.
This is an illustrative brief. Adapt it to a real task and use only approved material. It is deliberately ordinary: a reusable description of the work is more useful than a clever phrase that no one else on the team understands.
After review, note what the first draft misunderstood. Was the audience unclear? Was an important condition buried in the source? Did the requested format force a distinction to disappear? Change the brief based on that evidence.
Keep the final approved version and the revised instructions together. Over time, this gives the team a better working reference for a particular task. It does not guarantee correct output; it makes the task easier to communicate and check.
State the purpose, audience, approved sources, boundaries, output and review. Reuse the brief, then refine it from real corrections.
Canadian Centre for Cyber Security — Generative artificial intelligence ↗
For information-handling and verification considerations when using generative AI.
A practical way to find a useful AI starting point: follow one task, identify the friction, and define a result worth measuring.
How to design a small AI trial with an owner, a comparison, review criteria and a clear decision at the end.
A simple review routine for AI-assisted writing: check the facts, restore the context, and keep responsibility with the person approving the work.
Ideas, thoughtful disagreements, and new perspectives are welcome. Send a note.
[email protected]