I run AI training sessions for businesses across the UK, and the same question comes up in almost every one: “Where do I actually start with this stuff?”

It’s a fair question. AI tools have gone from novelty to necessity in a remarkably short time. But most professionals haven’t had any structured training on how to use them. They’re figuring it out on their own, picking up tips from LinkedIn posts and YouTube videos, and wondering whether they’re doing it right.

The truth is, you don’t need to become a data scientist or learn to code. But there are a handful of practical skills that make an enormous difference to how effectively you can use AI at work. I’ve seen it over and over: people who develop these skills get genuinely useful results from AI, while everyone else gets mediocre output and concludes the technology is overhyped.

Here are the five skills that matter most, with concrete examples of how they show up in day-to-day work. If you’re looking for a proper AI for business course or just want to upskill on your own, these are the capabilities worth focusing on.

1. Prompt engineering: getting AI to actually do what you want

“Prompt engineering” sounds technical, but it really just means: learning how to ask AI tools for things in a way that produces useful results.

Most people type something vague into ChatGPT, get a vague response, and give up. The skill is in being specific about what you need — the format, the audience, the tone, the constraints.

What this looks like in practice

Say you need to write a proposal email to a potential client. You could type “write me a proposal email” and get something generic and American-sounding. Or you could write:

“Draft a 200-word email to the operations director of a mid-sized logistics company. We’re proposing a three-month AI training programme for their management team. Tone: professional but warm, not salesy. Mention that we’ve worked with similar companies in the transport sector. End with a clear next step — suggesting a 20-minute call.”

The difference in output quality is night and day. The second prompt gives the AI enough context to produce something you can actually send, perhaps with a few tweaks rather than a complete rewrite.

The same principle applies to reports, summaries, data analysis requests — anything you ask AI to help with. The better your brief, the better the result. It’s the same logic as briefing a colleague: if you give someone vague instructions, you get vague work back.

How to build this skill

Start paying attention to which prompts give you good results and which don’t. Keep a note of the ones that work well — you’ll start to see patterns. The main things to specify are: the task, the format, the audience, the tone, any constraints (word count, structure), and relevant context. Our complete guide to AI for business covers prompt techniques in much more detail.

2. AI-assisted writing and editing: your drafting partner, not your ghostwriter

This is probably the most immediately useful AI skill for most office workers. Not because AI writes brilliantly — it often doesn’t — but because it’s remarkably good at getting you past the blank page.

The key mindset shift is treating AI as a drafting partner rather than a replacement for your own writing. You wouldn’t publish the first draft a junior colleague handed you without reviewing it. Same principle applies here.

What this looks like in practice

A marketing manager I trained last year described her process like this: she uses AI to generate a rough first draft of blog posts and email campaigns, then rewrites about 40% of it. She adds her own examples, adjusts the tone, cuts the waffle, and makes sure it actually sounds like her company. The whole process takes her about half the time it used to.

Another common use: editing and improving your own writing. You can paste a draft report into an AI tool and ask it to check for clarity, flag sections that are too long-winded, or suggest a better structure. It’s like having a second pair of eyes available whenever you need one.

Where people go wrong is using AI-generated text without any editing. Readers can usually tell. The writing tends to be technically correct but oddly flat — all the sentences are similar lengths, the vocabulary is a bit samey, and there’s a strange absence of personality. Good AI-assisted writing still has a human at the wheel.

How to build this skill

Practice the loop: generate a draft with AI, then edit it properly. Pay attention to what you consistently need to fix — that teaches you how to write better prompts next time. Try using AI for different stages of writing: brainstorming, outlining, drafting, editing, proofreading. You’ll find it’s more useful at some stages than others.

3. Data analysis with AI: making sense of numbers without a spreadsheet degree

This is the skill that surprises people the most. You don’t need to know pivot tables or VLOOKUP formulas to do meaningful data analysis anymore. You can upload a spreadsheet to an AI tool and ask questions about it in plain English.

I’m not suggesting you ditch Excel entirely. There are plenty of tasks where a proper spreadsheet is the right tool. But for quick analysis, pattern spotting, and making sense of unfamiliar datasets, AI is genuinely faster for most people.

What this looks like in practice

Imagine you’ve been handed a CSV file with 18 months of customer support tickets. Your manager wants to know which issues come up most often, whether response times have improved, and if there are any seasonal patterns.

In the old world, that’s a couple of hours of sorting, filtering, and building charts. With AI, you upload the file and ask: “What are the five most common ticket categories? How have average response times changed month by month? Are there any months with noticeably higher volume?” You get a summary with charts in a few minutes.

A finance director I worked with uses this approach for budget variance analysis. She uploads the monthly actuals alongside the budget, asks the AI to flag any line items that are more than 10% over or under, and gets a neat summary she can take straight into her board meeting. What used to take her team half a day now takes about twenty minutes.

How to build this skill

Start with a dataset you already know well — that way you can check whether the AI’s analysis is accurate. Ask simple questions first, then get more specific. Learn to be precise about what you’re asking for: “Show me a trend” is vague; “Compare Q1 and Q2 revenue by product category and highlight any categories where revenue dropped more than 15%” gives you something useful.

If you want structured practice with real business scenarios, that’s exactly what we cover in our AI training programmes.

4. Workflow automation: spotting the repetitive tasks AI can handle

Every job has tedious, repetitive tasks that eat up time. Formatting documents. Copying data between systems. Writing the same type of email over and over. Generating weekly status updates from the same sources.

The skill here isn’t necessarily building complex automations (though that’s valuable too). It’s learning to recognise which parts of your workflow are good candidates for AI assistance and which aren’t.

What this looks like in practice

An HR manager I trained realised she was spending roughly three hours every week writing personalised rejection emails to job applicants. The emails followed the same basic structure but needed to reference specific details from each application. She set up a simple workflow: paste the candidate’s name, the role they applied for, and one specific thing from their application into a prompt template, and the AI generates a thoughtful, personalised rejection email in seconds. She reviews each one before sending, but the time saving is substantial.

Another example: a project manager who receives weekly status updates from six different team leads in six different formats. He now pastes all six into an AI tool with the instruction to consolidate them into a single status report following his company’s standard template, flagging anything behind schedule. What took an hour now takes ten minutes.

The pattern is always the same: find a task that’s repetitive, follows a roughly predictable structure, and doesn’t require deep creative thinking. Those are your automation candidates.

How to build this skill

Spend a week keeping a log of everything you do at work. Flag anything you do more than twice that follows a similar pattern. Then ask yourself: could I explain this task clearly enough for someone else to do it? If yes, you can probably explain it clearly enough for AI to help with it. Start with low-stakes tasks where mistakes won’t cause problems, and build from there.

5. Critical evaluation: knowing when AI gets it wrong

This might be the most important skill on the list, and it’s the one people most often skip. AI tools are confident. They produce polished, authoritative-sounding text even when the content is completely wrong. If you don’t develop the habit of checking AI output, you will eventually publish, send, or present something inaccurate.

I’ve seen it happen more times than I’d like. A consultant who included AI-generated statistics in a client presentation, statistics that turned out to be fabricated. A solicitor whose AI-drafted legal summary cited cases that didn’t exist. An accountant whose AI-produced tax summary contained outdated allowance figures.

None of these people were careless. They just hadn’t developed the reflex of questioning AI output the same way they’d question any other source.

What this looks like in practice

Critical evaluation means different things depending on the task. For factual content, it means checking key claims against reliable sources before using them. For data analysis, it means sanity-checking the numbers — does this percentage actually make sense given what you know about the business? For written content, it means reading with a sceptical eye: is this accurate? Is anything missing? Would a knowledgeable reader spot problems with this?

It also means understanding AI’s known weaknesses. Large language models are prone to “hallucinating” facts and figures. They can reflect biases present in their training data. They sometimes produce text that sounds right but is subtly misleading. They’re not good at maths unless they’re using a specific calculation tool. They don’t know about very recent events unless they have access to current information.

None of this makes AI useless. But it does mean you need to stay in the driver’s seat. The professional who knows when to trust AI output and when to verify it is far more valuable than one who either blindly trusts everything or refuses to use AI at all.

How to build this skill

Make verification a habit, not an afterthought. Before you use any AI-generated content externally — in a report, presentation, email to a client — check the key facts. Ask yourself: what would go wrong if this were inaccurate? The higher the stakes, the more carefully you should check. Over time, you’ll develop an instinct for which types of AI output need close scrutiny and which are generally reliable.

Bringing it all together

These five skills aren’t independent of each other. Good prompt engineering makes AI-assisted writing better. Data analysis skills help you evaluate AI output critically. Understanding automation helps you apply all the other skills more efficiently.

The professionals who are getting the most value from AI right now aren’t the ones with the fanciest tools or the most technical backgrounds. They’re the ones who’ve taken the time to learn how to work with AI effectively — and that’s a skill set anyone can develop.

If your organisation is looking to build these capabilities across a team, we run practical AI training sessions designed specifically for business professionals. No jargon, no theory for theory’s sake, just the skills your team needs to work more effectively with AI tools.

You might also find our guide on AI training for business teams helpful if you’re thinking about how to roll this out across your organisation. And if you want a practical, step-by-step guide to getting started in your own role right now, our guide to using AI at work covers the daily habits and workflows that make these skills stick. For those in management roles, our guide to AI for managers covers how these same skills apply to team leadership and driving AI adoption across your team.

The tools will keep evolving. The skills to use them well? Those are worth investing in now.