A Universal AI Level. You’re Welcome, World.
- Sophie Boulderstone
- 2 days ago
- 5 min read
Apparently, nobody knows what level they are at with AI.
Nobody knows what level an AI course is aimed at either.
And employers keep writing things such as “must be confident using AI”, which could mean anything from asking ChatGPT to improve an email to building the machine that eventually replaces the finance department.
So, naturally, we fixed it.
You’re welcome, world.
The problem with beginner, intermediate and advanced
For years, courses have been divided into three extremely helpful categories:
Beginner.
Intermediate.
Advanced.
Lovely.
Except one person’s beginner course starts with, “Open ChatGPT.”
Another person’s beginner course starts with, “Connect the API, clone the repository and open the terminal.”
The same problem appears everywhere.
Someone says they “use AI”, but that might mean they chat to Claude when they are stuck.
It might mean they create images, spreadsheets and presentations.
It might mean they build automated workflows across several platforms.
Or it might mean they train foundational models and casually use the phrase “fairly straightforward” before drawing on six whiteboards.
Those people are not at the same level.
Yet we have had no widely understood way to describe the difference.
So we created one
The Universal AI Level, or UAIL, is a practical framework for describing AI ability from Level 0 to Level 10.
The key principle is simple:
Your UAIL is the highest level at which you can work independently and repeatedly.
Your level is not based on which tools you have opened.
It is not based on whether you know what an agent is.
It is not based on whether you once watched somebody build an automation on YouTube while nodding confidently. It is based on what you can actually do.
Can you complete the task yourself?
Can you do it again?
Can you check the result?
Can you recognise when it has gone wrong?
Can you make reasonable progress towards fixing it?
The Universal AI Levels
Level 0: Uninitiated
“I haven’t used AI.”
You have never intentionally used an AI chatbot, image generator or similar tool.
A perfectly valid place to begin. Everyone was here fairly recently.
Level 1: Conversational
“I talk to AI.”
You can hold a useful conversation with an AI assistant, provide context, ask for changes and improve the answer through discussion.
You have moved beyond typing one vague sentence, receiving something dreadful and declaring that AI does not work.
Level 2: Creator
“I use AI to make things.”
You can create images, documents, presentations, spreadsheets, audio, video or other useful outputs.
The important part is that you review and improve what it produces rather than copying the first thing it gives you and releasing it into the wild.
Level 3: Operator
“I use AI confidently in my work.”
You can use AI for research, analysis, planning, documents and data.
You choose appropriate tools, give useful context and check the result.
This is the important shift from making things with AI to properly working with it.
When AI confidently tells you something ridiculous, you investigate instead of adding it to the quarterly report.
Level 4: Automator
“I create repeatable AI workflows.”
You can build assistants, agents, templates and automations that complete recurring tasks.
You can design the process yourself and usually work out which part has broken when it inevitably decides to stop working five minutes before a demonstration.
Level 5: No-Code Builder
“I build connected AI systems without coding.”
You can create complete systems where AI tools, databases, connectors, agents and other services work together.
You may use MCP, APIs through visual tools, no-code platforms and AI-assisted interfaces.
You can build an actual functioning application without writing the underlying code yourself.
And yes, this is advanced AI capability, even if somebody on LinkedIn insists that it does not count because you did not personally type every bracket.
The code line
Levels 0–5 can be reached without coding.
Levels 6–10 involve code.
This is the dividing line that AI courses desperately need to make clear.
People should know whether “beginner” means learning how to communicate with an AI assistant or being introduced to Python by a man who says, “Don’t worry, this bit is easy.”
The coding levels
Level 6: AI-Assisted Coder
“AI helps me code.”
You can use AI to write and edit code, work with repositories, use a terminal, connect APIs and deploy a simple application.
You may still rely heavily on AI when things go wrong, but you can get a working product into the world.
Level 7: AI Developer
“I build and maintain AI applications.”
You can inspect, change, test and debug the code behind an AI application.
You understand enough about databases, authentication, architecture, security and deployment to maintain what you build.
When the AI writes something terrible, you can tell where the terrible bit is.
Level 8: AI Systems Engineer
“I architect production AI systems.”
You design reliable AI systems involving agents, retrieval, evaluations, monitoring, routing and orchestration.
You think about cost, security, speed, reliability and failure before customers find the problems for you.
Level 9: AI Architect or Researcher
“I build or adapt the underlying AI.”
You can train, fine-tune or substantially adapt models.
You understand machine learning, data, evaluation, optimisation, safety and infrastructure at a deep technical level.
Other developers build on the work you create.
Level 10: Creating the Matrix
“I created the AI.”
You develop foundational models, architectures, protocols or genuinely new approaches to artificial intelligence.
Other experts use your work to create their systems.
You are probably employed by an organisation with several billion pounds, a suspicious number of GPUs and no simple answer to the question, “So what exactly are you building?”
Why this is useful
Course creators can finally explain who a course is for.
Instead of saying:
Suitable for beginners and intermediate learners.
They can say:
Designed for UAIL Levels 2–3. No coding is required. By the end, learners should be able to work independently at UAIL Level 4.
That tells people what they need to know.
Employers can stop asking for candidates who are “proficient in AI” and explain the capability the role actually requires.
For example:
This role requires approximately UAIL Level 4. Candidates should be able to design and maintain repeatable AI workflows without coding.
Individuals can also describe their own ability more clearly.
Self-assessed UAIL Level 3, with experience in research, content development and data analysis.
Far more useful than “AI enthusiast”, which can currently mean almost anything.
It is free to use
The Universal AI Level is free to use, share and adapt.
There is no permission form.
There is no compulsory attribution.
We would obviously be delighted if people link back to the official version, particularly so others can find the complete definitions and future updates. But the point is for it to be useful and spread.
We needed a shared way to explain practical AI ability, so we made one.
View the complete Universal AI Level, download the guide and find your level at:
You’re welcome, world.
