Prompt engineering is the skill of writing instructions for an AI tool (Claude, ChatGPT, Copilot, Gemini, Grok) so that it gives a useful, accurate answer first time. It needs no code. In practice it means stating the role the AI should take, the task, the context it cannot guess, and the format you want back. We teach this as RICF.
Prompts tested in Claude and Grok, October 2026. Written by Rod Doyle, Director, TESS Group.
Key takeaways
- It is briefing, not coding. Tell the AI who to be, what to do, what it needs to know and what the answer should look like. Most good managers already brief people this way. They have not applied it to AI yet.
- RICF. Role, Instruction, Context, Format. You do not need all four every time, but when an answer disappoints, one of them is almost always missing.
- Four prompts you can copy. A difficult customer email, a long policy document, a research question and a page of messy meeting notes, each shown as the typical first attempt and the version that works. They run in Claude, Grok and most other assistants.
- Two lines that cut made-up answers. "If something is unclear, say so rather than guessing" and "do not invent missing details" do more than any clever phrasing.
- Safe prompting is part of good prompting. Approved tools, approved accounts, no personal data unless the setting is cleared for it, and check anything that matters against the source. Most teams are already using AI without anyone having set those rules.
This guide is for UK employers, managers and anyone in a non-technical role who uses, or is about to use, tools like Claude, Grok, ChatGPT, Copilot or Gemini at work. It keeps to the practical: what prompt engineering is, why it matters, a simple framework, and real before-and-after prompts you can copy.
Why it matters at work
Most organisations now have some AI tool in the building, licensed or not. What separates the teams getting real value from the ones quietly giving up is rarely the software. It is whether people know how to brief it.
Think of an AI assistant as a very quick, very well-read new starter who knows nothing about your business. Give them a one-line brief and you will get something generic. Give them the context, the goal and the format you need, and you will get something you can actually use.
That is prompt engineering. It is briefing, and most good managers already know how to do it. They just have not applied it to AI yet.
What is prompt engineering? The RICF framework
Tell the AI who to be, what to do, what it needs to know, and what the answer should look like. At TESS we teach this as RICF:
Role
Who should the AI act as? Example: "You are an experienced HR adviser in a UK SME."
Instruction
The actual task, as a clear verb. Draft, summarise, compare, check.
Context
The background it cannot guess. Audience, constraints, tone, what has already been tried.
Format
The shape of the output. A table, five bullets, a 150-word email, British English.
You do not need all four every time. But when an answer disappoints, one of them is almost always missing.
Prompt engineering examples: four prompts, before and after
Four everyday tasks, each shown as a typical first attempt and an improved version. They work in Claude, Grok and most other assistants.
1. Writing a difficult email (Claude)
Write an email to a customer about a late delivery.
You are a customer service lead at a UK B2B supplier. Draft a reply to a long-standing customer whose order is five working days late because our courier missed a collection. Be honest about the cause, do not over-apologise, and offer a firm new delivery date of Friday. Keep it under 120 words, British English, warm but direct. No promises about compensation.
The first version gets you something polite, long and bland, probably with American spelling and an apology that promises things you cannot deliver. The second gives the AI the relationship, the facts, the limits and the length. The draft needs a light edit, not a rewrite.
2. Summarising a long document (Claude)
Claude handles long documents well, so it is a good fit for policy packs, contracts and meeting notes.
Summarise this.
You are briefing a busy operations director who has two minutes. Summarise the attached policy in three parts: (1) what changes for frontline staff, (2) any deadlines, (3) anything that needs a decision from management. Use short bullets. If something is unclear in the document, say so rather than guessing.
That last line matters. Asking the AI to flag uncertainty is one of the simplest ways to reduce made-up answers.
3. Researching a topic (Grok)
Grok is useful when you want a quick read on something current. Be specific about what "good" looks like.
What is happening with apprenticeship funding?
I am an HR manager at a UK employer in England. Explain the main changes to apprenticeship funding for starts from 1 August 2026, in plain English. Split it into: levy payers, non-levy employers, and apprentices aged 16 to 24. Give me the source for each point so I can check it on GOV.UK. Keep it under 250 words.
Notice the request for sources. Any AI tool can get facts wrong, so always ask where an answer came from and check anything that matters against the original. (The right answers, if you want to mark its homework, are in our ST1512 funding FAQ.)
4. Turning messy notes into something usable (Grok or Claude)
Tidy up these notes.
Below are rough notes from a team meeting. Turn them into: a three-line summary, a table of actions with owner and due date, and a list of open questions. Do not invent owners or dates. If one is missing, write "TBC".
The "do not invent" instruction stops the AI filling gaps with confident nonsense.
Prompt engineering techniques: five habits that make the biggest difference
You can get most of the benefit from a handful of habits.
Give context, not just a task
The AI cannot see your organisation, your customer or your history. Tell it.
Say what format you want
Word counts, bullets, tables, headings. If you do not say, you get the AI's default, which is usually too long.
Show an example
If you have a previous email, report or post you liked, paste it in and say "match this tone and structure". Examples beat descriptions.
Iterate, do not restart
Treat the first answer as a draft. Reply with "shorter", "more direct", "turn this into a table", "what have you assumed?". A conversation gets you further than one perfect prompt.
Ask it to check itself
"What might be wrong with this?" or "What would a sceptical finance director challenge?" often surfaces the weak spots.
Anthropic's own guidance for Claude makes the same points: be clear and direct, use examples, and give the model room to think through a problem step by step before answering. The principles are not a TESS invention. They are what the people who build the tools recommend.
Prompt engineering glossary: six terms you will hear
You do not need the jargon to prompt well, but you will meet it in vendor documentation and training, so here is what it means in plain English.
- Zero-shot prompting
- Asking for a task with no examples. "Summarise this policy in three bullets." Fine for simple, common tasks.
- Few-shot prompting
- Giving one or more examples of the output you want before asking for a new one. Habit three above. The single most reliable way to get a consistent format or tone.
- Chain-of-thought prompting
- Asking the model to work through a problem step by step before giving its answer. Useful for anything with calculations, conditions or trade-offs.
- System prompt
- Standing instructions that apply to every message in a conversation or tool, such as "always answer in British English and cite sources". Business tools often let an administrator set one for the whole team.
- Context window
- How much text the model can hold in mind at once, including what you paste in. Long documents can exceed it, which is why summarising in sections sometimes works better.
- Hallucination
- A confident answer that is wrong or invented. Reduced, not removed, by asking for sources, telling the model not to guess, and checking anything that matters.
What prompt engineering is not
| The myth | The reality |
|---|---|
| It is coding | You write in plain English. The people who are best at it are often good communicators, not developers. |
| There are magic words | There is no secret phrase that unlocks hidden powers. Clarity beats cleverness every time. |
| It replaces judgement | A well-prompted AI still makes mistakes. You are responsible for what goes out with your name on it. |
| Learn it once | Tools change fast. The principles (clear role, clear task, real context, defined format) carry across, but the details shift. |
Staying safe while you prompt
Good prompting also means knowing what not to put in. Before anyone on your team pastes in customer records, staff data or anything commercially sensitive, they should know which tools your organisation has approved and what the rules are.
Use the AI tools your organisation has approved, on approved accounts. Do not paste personal data unless you know the tool and the setting are cleared for it. Check facts, figures and legal points against the original source. Be open about where AI helped; nobody should be quietly passing off unchecked AI output as their own.
This is where many teams get stuck. People are already using AI, often without telling anyone, and nobody has set the ground rules. If that sounds familiar, the 30-minute audit we published in September is the quickest way to find out what is in use.
From one good prompt to a team that works this way
Learning to write a better prompt takes an afternoon. Building it into how a team works takes longer.
That means knowing which tasks are worth handing to AI, how to chain prompts into a repeatable workflow, when to automate, and how to do all of that safely. It is the gap our AI and Automation Practitioner Level 4 apprenticeship is built to close. It is the Skills England standard ST1512, designed for non-technical staff in HR, finance, operations, marketing, sales and admin. Prompting, including the RICF framework, is covered in the first module, and apprentices go on to build working automations and AI agents in no-code tools, on real work in their own role.
Apprentices also take the NCFE AI Prompt Mastery qualification as standard, with other industry qualifications available. The full list is on the programme page, and our wider range of AI apprenticeships covers other levels. If what you need is a prompt engineering course for a team rather than an apprenticeship, our one-day AI Prompting Accelerator covers RICF and the habits above in a single session.
TESS Group has trained UK employers since 2009, is rated Good by Ofsted at its last inspection (June 2025), UKPRN 10031093, and delivers live online to employers across England.
Trying to get your team using AI properly, not just occasionally? Book 30 minutes with Rod. We will look at where your team is now, what is realistic, and whether an apprenticeship is the right route or something lighter would do.
Book a 30-minute call with Rod → or use the contact formThe honest summary
- What it is: briefing an AI clearly enough that it can act. Role, Instruction, Context, Format.
- What it is not: coding, magic words, or a substitute for your judgement.
- Where the value comes from: context and format, examples, iterating, and asking the tool to check itself.
- Where the risk comes from: pasting in data the tool is not cleared for, and trusting output nobody checked.
- The next step: one good prompt is an afternoon. A team that works this way is a programme.
Frequently asked questions.
What is prompt engineering in simple terms?
Prompt engineering is writing clear instructions for an AI tool so it gives you a useful answer first time. In practice, it means telling the AI what role to take, what task to do, what context it needs and what format you want back.
Do I need to know how to code to do prompt engineering?
No. Prompts are written in plain English. Good communicators who can brief a colleague clearly usually pick it up quickly.
Does prompt engineering work the same in Grok, Claude, ChatGPT and Copilot?
The core principles work across all of them: clear role, clear task, real context and a defined format. Each tool has its own strengths, so it is worth testing the same prompt in more than one and seeing which suits the job.
How do I stop AI making things up?
You cannot stop it completely, but you can reduce it. Ask the AI to say when it is unsure, tell it not to invent missing details, ask for sources, and always check anything important against the original.
Is prompt engineering a real job skill or a fad?
It is a real workplace skill. The term may change, but the ability to brief AI tools clearly, check their output and build them into everyday work is becoming part of many roles.
What is the RICF prompt framework?
RICF is the four-part structure TESS teaches for writing a prompt: Role (who the AI should act as), Instruction (the task, as a clear verb), Context (the background it cannot guess, such as audience, constraints and tone) and Format (the shape of the answer, such as a table, five bullets or a 150-word email in British English). You do not need all four every time, but a disappointing answer is usually missing one of them.
What is the difference between zero-shot and few-shot prompting?
Zero-shot prompting asks for a task with no examples. Few-shot prompting gives the AI one or more examples of the output you want before asking for a new one. Few-shot is the more reliable way to get a consistent format or tone, which is why "paste in an example and say match this" is one of the five habits in this guide.
How can my staff learn prompt engineering properly?
Short practice helps, but the deepest learning comes from using it on real work over time. Our AI and Automation Practitioner Level 4 apprenticeship builds prompting into a wider programme of practical AI and automation skills, with no coding required.
Rod Doyle is a Director of TESS Group, an Ofsted-rated Good apprenticeship provider founded in 2009 that delivers the AI and Automation Practitioner Level 4 (ST1512) and AI leadership units live online across England. He writes the employer guides on this site. LinkedIn.
Sources: Anthropic, prompt engineering overview (be clear and direct, use examples, let the model think step by step); GOV.UK, Apprenticeship funding, policy for starts from 1 August 2026, used for the research example; TESS Group programme page for the RICF framework, module order and the NCFE AI Prompt Mastery qualification. The before-and-after prompts are our own and were tested in Claude and Grok in October 2026. TESS Group delivers AI apprenticeships and has a commercial interest in employers choosing structured training.