Most people prompt AI the way they'd type into a search box: a few words and hope. Then they decide the tool is rubbish. The gap between a vague prompt and a good one is usually the gap between a useless answer and an hour saved.

I've spent 23 years in marketing and I've reviewed more than three hundred tools on this site. Writing clear prompts is the same skill as writing a clear brief for a freelancer, and that's how I teach it.

What business teams get wrong about prompts

The problem isn't that people don't know clever tricks. It's that they leave out the things a new colleague would need: who the work is for, what good looks like, what to avoid and what format to hand back. Prompt engineering for a business team is mostly about putting that context in, every time, without having to think about it.

• No context: asking for a proposal without saying who it's for or what you sell.
• No examples: describing the style you want instead of showing one.
• No format: getting a wall of text when you needed a table or three bullet points.
• Starting from scratch: rewriting the same prompt every week instead of saving it.

What OpenAI's own guide says that most courses skip

OpenAI's prompt engineering guide for its API is written for developers, but two of its points matter just as much to a marketing team. The first is few-shot learning: include a handful of input and output examples in the prompt, and show a range of possible inputs with the outputs you want. One good example of your house style beats a paragraph describing it.

The second is testing. OpenAI recommends building tests and evaluation suites that measure prompt behaviour so you can monitor performance as you iterate, and pinning production apps to a specific model snapshot so behaviour stays consistent. In a small business that translates into something simple: keep a few real test cases for each shared prompt and rerun them when the tool updates, because a prompt that worked last month can drift.

How the training runs

Every session uses your team's real work, so the prompts they leave with are ready to use on Monday.

1. Each person brings two or three tasks they repeat every week.
2. We rewrite their prompts together with context, examples and a clear output format.
3. We test each prompt on a few real cases and fix what breaks.
4. The best prompts go into a shared library the whole team can use.
5. We agree who owns the library and how often it gets checked.

A prompt structure your team can remember

I teach one simple structure that works in ChatGPT, Claude, Gemini and Copilot. Say who the AI is helping and why. Give it the material to work from. Show an example of a good result. Say exactly what to hand back, in what format and length. Then say what to leave out.

It sounds obvious written down. In practice almost nobody does all five, and the difference shows straight away. Once the team uses the same structure, prompts become easy to share and easy to fix, because everyone can see which part is missing.

Who this suits

It suits marketing, sales, operations and customer service teams who already use AI tools but get patchy results. It suits owners who want their team using AI the same way instead of twelve different ways. You don't need any technical background. It runs over video and in English, for teams in the UK, US, Canada, Australia and beyond.

If you'd like a head start before the session, my prompt engineering cheat sheet covers the basics, and ChatGPT training for teams goes wider than prompts. Every way to work together is on my work with me page.

Prompts for the jobs teams repeat most

Most teams get the quickest win from the same handful of tasks. Turning meeting notes into actions and owners. Drafting a reply to a common customer question in your tone. Summarising a long document into the five points a manager needs. Turning a rough idea into a first draft of a newsletter or a proposal.

For each of those we build one shared prompt with the context, an example and the format baked in. The person using it only fills in the new material. That's what makes AI use consistent across a team: not everyone getting clever with prompts, but everyone starting from the same good one.

I also show people how to ask the AI to check its own work, for example by listing anything it wasn't sure about, and how to spot the moment when it's better to stop prompting and write the thing yourself.

My plain advice on prompt engineering for a team

Don't chase magic phrases. The prompts that work in a business are the boring ones that include the context, an example and a clear format. Save them, name an owner, and test them when the tools change.

And measure the result, not the prompt. If a saved prompt turns a two-hour job into twenty minutes with a quick human check, it's a good prompt, however plain it looks.

Frequently asked questions

Who offers prompt engineering training for business teams?

Lilach Bullock, a British AI trainer with 23 years in marketing, runs prompt engineering training over video for marketing, sales and operations teams, built on their real tasks.

What is few-shot prompting?

It means including a handful of examples of the input and the output you want inside the prompt. OpenAI's guide recommends showing a range of inputs with the desired outputs.

Why do prompts stop working?

AI tools update their models, and a prompt that worked before can drift. OpenAI recommends tests that measure prompt behaviour, so keep a few real test cases and rerun them.

Do business teams need to learn coding for prompt engineering?

No. Business prompt engineering is about clear context, examples and output format. It's closer to writing a good brief than to programming.

What should a shared prompt library include?

The prompt itself, an example of a good result, who owns it, and a few test cases to check it still works after a tool update.