What is a large language model? A plain-English guide
How large language models like the ones behind ChatGPT, Claude and Gemini actually work, what they're good at, where they fall short and how businesses use them.
How to write prompts that get consistent, useful results from AI models, with reusable patterns, examples and the habits that separate good prompts from guesswork.
Good prompts tell the model who it's helping, what the task is, what information to use, what the output should look like and what to avoid, often with an example. Treat prompts like reusable templates: write them carefully, test them on real cases, keep the best versions and share them across your team. Clear inputs produce reliable outputs.
"Prompt engineering" sounds technical, but most of it is clear communication. The same habits that help you brief a new colleague help you brief an AI model. Here's a practical approach for teams that want consistent results rather than lucky ones.
A language model only knows what's in its training and what you put in front of it. A one-line request leaves it to guess your audience, tone, length and purpose. A well-structured prompt removes the guessing. Our guide to what a large language model is explains why the model fills gaps the way it does.
1. Role and audience. "You're helping a small accountancy firm reply to client emails. Clients are busy business owners, not accountants."
2. Task. State exactly what you want: "Draft a reply confirming we received their documents and explaining the next two steps."
3. Context. Paste the relevant facts: the client's email, your policy, dates, prices. The model can't use information it doesn't have.
4. Constraints. Length, tone, words to avoid, things not to promise: "Under 120 words. Friendly but professional. Don't mention fees."
5. Output format. Bullet points, a table, JSON for software, headings, a subject line plus body.
6. Examples. Show one or two examples of good output. This is often the single most effective improvement.
Many teams adopt a simple structure like this:
Save templates for repeated tasks, such as proposal sections, customer replies, product descriptions and meeting summaries, so everyone starts from a tested version.
Ask for a plan first. For complex tasks, ask the model to outline its approach before writing the full answer. You can correct the plan cheaply.
Break big tasks into steps. Summarise first, then analyse, then draft. Chaining smaller prompts often beats one giant prompt.
Tell it what to do when it doesn't know. "If the information isn't in the text provided, say so rather than guessing." This reduces invented answers; see our guide to AI hallucinations.
Use delimiters. Mark pasted content clearly, for example with headings such as "Customer email:" and "Policy:", so the model knows what's instruction and what's material.
Request specific evidence. Ask the model to quote or reference the part of your document that supports each point.
Iterate. Treat the first answer as a draft. "Shorter." "Simpler words." "Add a clear next step." Small follow-ups are fast and effective.
When AI is built into an application, prompts become part of the product. They need the same care as code:
If you find yourself pasting the same documents into every prompt, it may be time for a system that retrieves your information automatically. Our AI solutions team builds assistants connected to company knowledge, our AI chatbot team creates customer-facing bots with tested prompts and guardrails, and our web app team builds AI features into your own software.
Yes, although much of it is clear writing and structured thinking. Good prompts make AI outputs more accurate, consistent and useful.
Mostly, but each model has quirks. Test important prompts on the model you'll actually use.
As long as it needs to be to give the model the context, rules and format it needs. Detailed prompts often outperform short ones.
This guide is general information, not legal, tax or financial advice, and details such as prices, features and rules can change. Product and company names mentioned are trademarks of their respective owners and are used only to identify those products; UK Technology Hub is not affiliated with or endorsed by them.
Tell us about your project and get a free, fixed quote.
Get a free quoteHow large language models like the ones behind ChatGPT, Claude and Gemini actually work, what they're good at, where they fall short and how businesses use them.
A practical approach to testing AI features: building test sets, measuring accuracy and safety, using automated checks and monitoring quality after launch.
What AI hallucinations are, why language models produce confident errors, and the practical techniques businesses use to keep answers accurate.