The AI competency pyramid.
Where 20.6 million UK workers actually sit — 14 distinct levels, from the people who’ve never knowingly used AI to the few hundred who build the models.
How deep do you want to go? Pick a level and the article rewrites itself.
According to the ONS, 34.31 million people are in employment in the UK. An estimated 20.6 million use a computer every day. That’s the entire addressable market for AI at work — and it splits across 14 levels of competency, from people who’ve never knowingly used it to the few hundred who build the models themselves.
I mapped every level. The distribution is uncomfortable: 95% of UK computer users sit in the bottom four levels. Everything above Level 4 combined adds up to less than 5%.
Levels 0–4: the general workforce.
Five levels covering 98% of the country’s computer users. If you work in an office and send emails for a living, you’re almost certainly here.
Level 0 — Non-user 25%
Doesn’t use AI deliberately. Meets it through autocomplete or spam filters, but ignores or distrusts it.
Level 1 — Passive consumer 40%
Uses AI baked into everyday tools without calling it “AI” — accepts Copilot’s email suggestions, lets Gemini summarise a thread. The single largest band, and where most of the country lives.
Level 2 — Casual questioner 20%
Uses ChatGPT, Claude or Gemini as a better search engine. Quick, throwaway, single-turn prompts.
Level 3 — Practical user 10%
Feeds documents into AI for specific tasks: proofreading, summarising a PDF, drafting to an audience. Treats AI as a capable assistant.
Level 4 — Applied reasoner 3%
The mindset shift. Uses AI to compare and reconcile across sources, iterates on prompts, treats it as a collaborator rather than a tool.
The jump from Level 3 to Level 4 is free. No budget, no platform, no IT approval — just a change in how you approach the machine. Only 3% have made it.
Levels 5–9: builders to frontier.
Everything above the general workforce — under 5% of computer users combined. Where AI stops being an assistant and starts being infrastructure.
Level 5 — Builders (~1.6%). Automation builders chain AI into workflows with n8n, Zapier or Power Automate; integration engineers wire it to enterprise systems via APIs and MCP; security practitioners assess it for data-leakage and compliance risk.
Level 6 — Operators (~0.25%). Deploy agents that plan and act semi-autonomously — plus the red teamers who try to break them. The jump here is cultural, not technical: tolerating non-determinism.
Level 7 — Architects (~0.1%). Design AI-native platforms: retrieval pipelines, fine-tuned models, evaluation frameworks.
Levels 8–9 — Frontier (<10k people). Safety engineers, frontier problem-solvers and model researchers. Most of the UK’s Level 9 cohort would fit in a single townhall.
What the numbers tell us.
95% at Levels 0–3 is 19.57 million people — more than London, Birmingham, Manchester and Leeds combined — getting little from AI beyond autocomplete and summarisation.
Three things stand out. The 3→4 jump is free and almost nobody takes it. The 5→6 jump is where enterprise transformation projects die — on culture, not technology. And the safety cohort is dangerously small: roughly three safety-focused people for every ten thousand deploying AI.
The strategic question.
Which level is your organisation operating at, and which level do you need it at in two years? For most UK enterprises the honest answer is “Level 2–3, and we need Level 5 by 2027.” That’s a two-level leap across a whole workforce — a mindset, infrastructure and culture problem at once. The winners won’t buy the most tooling; they’ll move their median employee one level up the pyramid every year.
Move your median employee one level up the pyramid every year. That is the game.
More issues live on Substack.
Practical, evidence-led writing on AI adoption, capability and agentic workflows. Subscribe and each new issue lands in your inbox — no hype.