The short answer
An AI skills assessment measures how well someone uses AI tools at work, not how much they know about AI. This free one scores eight practical skills from 0 to 4, totals them out of 32 and places you in one of five bands. It runs in your browser, the method is published below and there's no email wall.
Most AI assessments online are either a personality quiz with a sales call attached, or a test of whether you can define "large language model". Neither tells you much about whether someone can get a proposal written in half the time without leaking a client's bank details.
So I built the one I wanted when scoping training for teams. Eight skills, described as things you actually do. A score you can work out on the back of a napkin. A safety rule that stops a reckless power user being rated "advanced". And a team mode, so a manager can see where the gaps really are before spending a penny. Ten minutes, honest answers, no email address required.
Take the free AI skills assessment
Your team
| Person | Prompting | Checking | Safe use | Tools | Tasks | Data | Automation | Habit | Total | Band | Remove |
|---|
How does the AI skills assessment scoring work?
Each of eight skills is scored 0 to 4 using plain descriptions of what people actually do. The scores add up to a total out of 32, which places you in a band from Grounded to Flying solo. One rule overrides the total: score 0 or 1 on safe and responsible use and you can't rate above Taxiing.
A few deliberate choices, and why:-
- Behaviour, not knowledge. Every level describes something you do. "I check figures before anything leaves my desk" is checkable. "I understand AI accuracy" isn't.
- Equal weights. I could weight prompting double and safety triple, but I'd be inventing precision I don't have. Equal weights are simple to explain and hard to fiddle. The safety rule does the one job weighting would do.
- The safety cap. Someone who drafts brilliant reports by pasting client contracts into a personal ChatGPT account isn't advanced. They're a data incident with good formatting.
- Habit counts. Skills you use once a quarter fade. Frequency is the best single predictor of whether training sticks, so it gets a line of its own.
- Nothing leaves your browser. The maths runs on the page. If you add people to a team, their scores are stored in your browser only, on your device.
What are the eight AI skills this assessment measures?
The assessment measures prompting and briefing, checking the output, safe and responsible use, choosing the right tool, everyday work tasks, documents and data, automation and agents, and habit. Together they cover the skills most office workers need to use AI tools productively and safely in 2026.
Foundation skills: how you work with AI
Prompting and briefing
Telling the tool who it is, what you need and what good looks like. The difference between level 1 and level 3 is usually twenty extra words and a huge jump in quality. Our prompting masterclass goes deep on this.
Checking the output
AI is fluent, fast and occasionally makes things up with total confidence. Checking is the skill that separates useful from dangerous. See how to judge the accuracy of AI information.
Safe and responsible use
Knowing what you can put in, which accounts are approved, and what the settings do with your data. It's the skill with the cap attached, for good reason. Start with keeping information safe when using AI tools.
Choosing the right tool
ChatGPT, Claude, Copilot, Gemini and Perplexity are good at different things. Using one for everything is like doing all your DIY with a butter knife. Here's the stack I actually pay for.
Applied skills: what you get done with AI
Everyday work tasks
Emails, drafts, summaries, meeting notes, proposals. This is where most people win back their first few hours a week. More in how to use AI for work tasks.
Documents and data
Pulling the point out of a forty-page document, explaining a spreadsheet, spotting a trend. If your world is Excel, start with Copilot prompts for Excel.
Automation and agents
Getting AI to do repeat or multi-step jobs with less hand-holding. Most teams don't need to start here, but somebody on the team should know what's possible. See AI agent training for non-technical teams.
Habit
How often you actually use it. Not a skill in the textbook sense, but nothing else on this list survives without it.
| Skill | Level 1 looks like | Level 2 looks like | Level 4 looks like |
|---|---|---|---|
| Prompting and briefing | I type short questions, like a search engine, and take what comes back. | I give some context and ask for a format, and I'll ask for another go if the first one's off. | I keep reusable prompts or custom assistants for repeat jobs, and colleagues borrow them. |
| Checking the output | I skim it and fix anything that looks obviously wrong. | I check the facts, figures and names that matter before anything leaves my desk. | I have a checking routine for each type of output and can spot a confident hallucination at fifty paces. |
| Safe and responsible use | I paste in whatever helps, client and personal details included, into whichever tool is open. | I keep obvious personal and confidential data out, but I'm hazy on which tools and settings are approved. | I could explain the risks to a colleague, and I help write or update our rules. |
| Choosing the right tool | I use one tool for everything because it's the one I know. | I've tried a couple and have a rough idea what each is good at. | I test tools on new tasks, keep up with changes and could recommend a setup for my team. |
| Everyday work tasks | I've used it once or twice for a one-off, like an awkward email. | I use it most weeks for drafting or summarising. | I've redesigned parts of my job around it and can show you the hours it saves. |
| Documents and data | I've asked it to summarise a document. | I use it to pull key points from long documents or explain a spreadsheet. | I combine several files or sources into reports or decisions other people rely on. |
| Automation and agents | I know roughly what automation and agents are, but haven't tried either. | I've set up something simple and repeatable, like a template, saved workflow or custom GPT. | I build and look after automations or agents other people use, and I know where they need a human. |
| Habit | A few times a month at most. | About once a week. | Several times a day. I'd notice if it vanished. |
What do the AI skills score bands mean?
Scores of 0 to 7 are Grounded, 8 to 15 Taxiing, 16 to 23 Take-off, 24 to 28 Cruising and 29 to 32 Flying solo. Most people starting out land in Taxiing: they've tried AI but don't use it routinely. The bands come with a suggested next step rather than a grade.
| Band | Score out of 32 | What it usually means | Sensible next step |
|---|---|---|---|
| Grounded | 0 to 7 | You haven't really started yet, which is fine, and more common than LinkedIn would have you believe. | Do a free AI Skills Boost course, then get hands-on with Intro To AI. |
| Taxiing | 8 to 15 | You've tried AI, but it isn't part of how you work. Most people are here. | A practical session on your own tasks: Intro To AI (90 minutes) closes this gap fastest. |
| Take-off | 16 to 23 | You use AI most weeks and get real value, with gaps in checking, data or automation. | Fill the gaps with the half-day AI Basics Deep Dive. |
| Cruising | 24 to 28 | AI is a regular part of your working week and you use it with care. | Go deeper on one area: Copilot, GEO or a specialist module. |
| Flying solo | 29 to 32 | You're a confident power user and probably the person colleagues already ask. | Build agents with the Claude Code pathway, and help bring your team up with you. |
Yes, the bands are named after flying. We're called The Oxford AI School and our motto is "Learn To Fly With AI", so it was that or naming them after Hobnobs.
How do I assess my whole team's AI skills?
Ask each person to complete the assessment, either on their own device or with you, and add each result to the team view. The page averages every skill, flags the weakest skill as your training priority, and warns you where the team is uneven or where anyone sits below the safety line.
- Send people the link and ask for honest answers, not flattering ones. Tell them it isn't a performance review, because it isn't.
- Have each person read their total and band out to you, or sit with them and enter it yourself using "Add to team".
- Look at the weakest skill on average. That's where your first training session should go.
- Look for spread. If one person scores 4 on a skill and another scores 0, pair them up before paying anyone.
- Deal with anyone below the safety line first, whatever else you do.
The team view lives in your browser on this device, so it's private but not shared. If you'd like us to run a fuller benchmark across a whole business, our AI capability assessment does that, and the AI skills matrix template gives you a spreadsheet version. For the organisation-wide picture (policy, tools, data and leadership) use the AI readiness assessment.
How does this compare with the Skills England AI benchmark?
Skills England's AI foundation skills for work benchmark covers understanding risks, creating prompts and adapting settings, using simple AI dashboards to read data trends, and automating routine tasks. This assessment covers all four, then adds checking output, choosing tools, everyday work tasks and habit, which decide whether training actually sticks.
| Skills England benchmark area | Skill in this assessment |
|---|---|
| Understanding the risks and consequences of using AI | Safe and responsible use (with the safety cap) |
| Creating prompts and adapting settings | Prompting and briefing; Choosing the right tool |
| Using simple AI dashboards to interpret data trends | Documents and data |
| Automating routine tasks with AI tools | Automation and agents |
| Not a separate benchmark area | Checking the output; Everyday work tasks; Habit |
The free government courses under AI Skills Boost are badged against that benchmark, so if you score Grounded or low Taxiing, one of those is a perfectly good first step. They're listed in our guide to free AI training courses in the UK.
Why do people overrate their AI skills?
Most people teach themselves AI by trial and error, so they don't know what they're missing. A UK survey for The Access Group found 70% of workers who use AI learn by experimenting, while only 19% had taken a course. Self-scores tend to run high on prompting and low on safety awareness.
That Access Group research matches what I see in every room I teach. People are often genuinely good at getting a decent first draft. They're much less sure where the data goes, and they rarely check citations. The fix is cheap: score yourself, ask a colleague who's seen your work to score you too, and take the lower number. It stings for about a minute.
How often should you reassess? Every three months, or after any training. Scores move fast once people use AI daily, and seeing the jump is half the motivation.
Want help closing the gaps?
Send me your team's scores. I'll tell you which skill to fix first and whether that takes a 90-minute session, a half day or nothing more than a decent policy and a fortnight's practice.
What should I do with my AI skills score?
Fix the lowest-scoring skill first, not the band. If safe use is 0 or 1, sort that before anything else. Then practise one skill on real work for a fortnight and reassess. Training helps most when it targets your weakest skill on your own tasks.
If you're weighing up paid training, read the AI training buyer's checklist before you book anyone (us included), and use the AI training price guide to sanity-check quotes.
Key takeaways
- A good AI skills assessment measures what people do with AI at work, not what they know about it.
- This one scores eight skills from 0 to 4, for a total out of 32 and five bands from Grounded to Flying solo.
- Scoring 0 or 1 on safe and responsible use caps you at Taxiing, whatever your total.
- It covers all four areas of Skills England's AI foundation skills benchmark, plus checking, tool choice, everyday tasks and habit.
- Self-assessments run high, so score yourself, ask a colleague, and take the lower number.
- For teams, train the weakest average skill first and fix anyone below the safety line straight away.
Frequently asked questions
What is an AI skills assessment?
An AI skills assessment measures how well a person uses AI tools at work: how they brief the tool, check its output, keep data safe, pick the right tool and fit AI into everyday tasks. Good ones score behaviour you can observe, not knowledge of AI theory.
How long does the AI skills assessment take?
About ten minutes for one person. There are eight questions, each with five statements to choose from. A team of eight takes most managers under an hour if people complete it themselves and share their scores.
Is this AI skills test free?
Yes. It's free, needs no sign-up and doesn't ask for an email address. The scoring runs in your browser and nothing you enter is sent to The Oxford AI School.
How is the AI skills score calculated?
Eight skills are scored 0 to 4 and added up for a total out of 32. Totals of 0 to 7 are Grounded, 8 to 15 Taxiing, 16 to 23 Take-off, 24 to 28 Cruising and 29 to 32 Flying solo. Anyone scoring 0 or 1 on safe and responsible use is capped at Taxiing.
What is a good AI skills score?
For most office roles, 16 or more (Take-off) means AI is genuinely useful in your week, and 24 or more (Cruising) means you use it routinely and carefully. A balanced profile with at least 2 in every skill matters more than a high total.
Can I use this to assess my team's AI skills?
Yes. Add each person's result to the team view to see averages per skill, the weakest skill to train first, uneven skills and anyone below the safety line. For a fuller benchmark, see our AI capability assessment.
Is my data stored anywhere?
Only in your own browser on the device you used, so you can come back to it. Nothing is sent to us or anyone else. Clearing your browser data, or pressing Clear team, removes it.
How does this relate to the government's AI Skills Boost?
Skills England's AI foundation skills benchmark, used to badge the free AI Skills Boost courses, covers risks, prompting, data and automation. This assessment covers those four and adds checking output, choosing tools, everyday tasks and habit.
About Harry Lang
Harry runs The Oxford AI School in Witney, Oxfordshire, teaching UK teams the basics of AI in a fun, easy and practical way. He's a former CMO who has scaled B2C brands globally, and the author of Learn To Fly With AI and Brands, Bandwagons & Bullshit.
Last reviewed 2 November 2026. Prices and figures change; if you spot one that has moved, email Learn@TheOxfordAISchool.com and I'll fix it.