Prompting for business users: a practical guide

Clear instructions produce better results. That is true when briefing a colleague, and it is just as true when asking an AI assistant to draft an email, summarise a report or prepare meeting notes. Most disappointing output comes from instructions that were too short or too vague.

Business users do not need to learn technical tricks. A few habits, practised until they are automatic, make most of the difference. This checklist sets them out.

State the goal and audience

Say what you want and who it is for. ‘Write an email about the price change’ leaves the assistant guessing. ‘Write a short email to our existing trade customers explaining a 4% price change from 1 March, in a friendly but direct tone’ gives it what it needs.

• What is the task: draft, summarise, rewrite, compare or list?
• Who will read the result: a customer, the board, a new employee?
• What should they do or understand after reading it?
• How long should it be, and in what format: email, bullet points, table?

Give context and examples

Provide the facts and a sample of the style you need. The assistant does not know your products, your customers or the history behind a decision unless you tell it. Paste in the relevant facts, or point it to approved documents if your tool allows that.

An example is often worth more than a paragraph of description. If you want a summary in the style of last month’s board update, include a short extract from it. If you want a reply in your company’s tone, include a previous reply you liked.

Review and refine

Check the output and ask for changes rather than starting again. The first draft is a starting point. Read it as you would a junior colleague’s draft: is it accurate, is the tone right, has anything been invented?

Then ask for specific changes. ‘Make it shorter’ works. ‘Cut it to three paragraphs and move the deadline to the first sentence’ works better. Building on a draft is usually quicker than rewriting the original instruction from scratch, because the assistant keeps the context of the conversation.

Check facts before anything leaves your desk

AI assistants can produce confident, well-written text that contains errors: a wrong date, a figure that was never in the source, a policy that does not exist. You remain responsible for anything you send. Check names, numbers, dates and commitments against the source every time.

A worked example

Consider an operations manager preparing a monthly update for the leadership team. A weak instruction would be ‘summarise this report’. A stronger one would be: ‘Summarise the attached operations report for the leadership team in five bullet points. Lead with anything that needs a decision. Keep each point to two sentences and use the same plain style as the example below.’ The manager then pastes last month’s update as the example.

The first draft is close but buries a supplier delay in point four. The manager asks for it to move to the top with the expected impact on deliveries. Two short exchanges produce an update ready for a final read, in a fraction of the time a blank page would take.

A prompting checklist

Build a small prompt library

Once a prompt works well, save it. Teams that share a handful of tested prompts for their common tasks, such as weekly summaries, customer replies or meeting notes, get more consistent results and spend less time experimenting. Keep the library short, give each prompt an owner and update it when the tool or the task changes.

Try it this week: pick one task you do repeatedly, write a prompt using the checklist above, refine it twice and save the version that works.

Source: Intelliversal analysis.

Frequently asked questions

Do I need special training to write good prompts?

No. The habits that make a good brief for a colleague, a clear goal, audience, context and an example, make a good prompt. Short practice sessions on real tasks help more than theory.

Why does the same prompt give different answers?

AI assistants vary their output slightly each time. That is why reviewing the result matters, and why saved prompts with clear instructions and examples give more consistent results.

Can I paste customer information into an AI tool?

Only if your organisation's policy allows it for that specific tool. Many approved business tools have different rules from public ones. If you are unsure, ask your manager or IT before pasting.

Explore Generative AI, copilots and agentic automation