Prompt vs Task specification

A prompt is the text you send to the AI assistant, such as a question or instruction. A task specification is a detailed outline of what you want: the outcome, inputs, rules, process, and how you will judge if the answer is good enough.

If you only send a brief prompt, the AI may not know what is important to you. A task specification—listing what you want, what matters, and any limits—helps the AI focus on the right details.

1. Decide what outcome you want.
2. List any documents, data, or information the AI should use.
3. Set rules or constraints (like word count or style).
4. Describe the steps you want the AI to follow.
5. State how you will judge if the answer is good enough.

Yes, anyone can use a task specification. You do not need to code or use technical language. A task specification is about being clear on what you want, not about programming.

Not always. Sometimes you can include all the details in your prompt, but for complex tasks, it helps to plan your task specification first—then turn it into a clear prompt.

The AI will do its best with what you send, but it may miss key points or misunderstand what you want. A prompt alone often leads to vague or incomplete answers, especially for bigger tasks.

1. Read your prompt and ask: does it say what you want?
2. Check if it lists any rules or limits.
3. See if it mentions how you will judge the answer.
4. If anything is missing, add it before sending.

Yes. A clear task specification helps the AI focus, use the right sources, and deliver an answer that matches what you need. It reduces the risk of missed details or wasted time.

• Desired length or word count
• Key points or sections to focus on
• Deadline for the summary
• Any style or tone requirements
• How you will check if the summary is good enough

No. A prompt can be a question, a command, or even a fragment. What matters is that it is clear enough for the AI to understand what you want.

For simple questions, a prompt is enough. For bigger or more important jobs—like reports, summaries, or anything with risk—a task specification helps you get what you need.

Yes, and it often helps. Writing out your task specification lets you check that you have covered all the details before turning it into a prompt.

• Leaving out key details
• Not stating what counts as a good answer
• Forgetting deadlines or limits
• Assuming the AI knows what matters most

A task specification is your plan for the job. The prompt is how you deliver that plan to the AI. For simple tasks, they may be the same. For complex ones, the prompt should reflect your full task specification.

Yes. If you often ask for the same kind of work—like summarising contracts or checking grammar—you can reuse your task specification and adapt the prompt each time.

An acceptance criterion is a rule or standard for judging if the AI's answer is good enough. For example, 'summary covers all main clauses and is under 200 words' is an acceptance criterion.

As detailed as needed for the job. For a quick question, a short prompt works. For a big task, include every rule, source, and check you care about. More detail usually means better results.

No. A task specification helps you get a better answer, but you still need to check the result—especially for important or risky work.

You can, but writing it down helps you spot gaps and makes it easier to copy into the AI prompt. Written specifications are easier to reuse and share.

If you skip the task specification, the AI may miss important points, misunderstand your needs, or deliver something you cannot use. For big jobs, a task specification saves time and reduces mistakes.

1. List your outcome, inputs, rules, and checks.
2. Write a clear instruction combining these points.
3. Paste it into the AI assistant.
4. Check the answer against your original specification.