The 10 Levels of AI Thinking: Where Are You?
By John R. Kennedy ·
Most people are learning how to use artificial intelligence. Far fewer are thinking about what happens when it changes everything around them.
I have conversations about artificial intelligence almost every day. Some are with business owners, some with technology professionals, and others with people who have only recently started experimenting with ChatGPT.
I've noticed something interesting about these conversations. Most revolve around what AI can do today.
Someone discovered how to write a better email. Another person figured out how to analyze a spreadsheet. Someone else built an application without knowing how to write a single line of code.
These are remarkable accomplishments. Just a few years ago, many of them would have been unimaginable.
But I think we're missing something important.
We're becoming increasingly proficient at using a technology without necessarily understanding the world that technology is creating.
It's a little like learning how to drive a car without considering how automobiles would eventually change cities, commerce, employment, and even where people chose to live. The automobile wasn't simply a faster horse. It fundamentally changed how society functioned.
I believe artificial intelligence will have an even greater impact. Yet most of our conversations remain focused on what we can accomplish with it right now.
That observation led me to develop a framework I call the AI Vision Spectrum.
It's a way of describing the different levels at which people think about artificial intelligence. The framework isn't a scientific classification or a measure of intelligence. It's simply a way to recognize how far beyond the immediate application of AI someone is looking.
A brilliant physician might have little interest in the infrastructure supporting AI, while a software architect might spend years thinking about it. Neither is necessarily more intelligent than the other. They're simply approaching the technology from different perspectives.
And nobody occupies one level all the time.
What interests me is how our thinking changes as we begin to understand the larger picture.
From Using AI to Building With It
At the beginning of the spectrum are AI Users, Levels 0 through 2.
These are people discovering what artificial intelligence can do for them. They ask questions, generate images, summarize documents, write emails, and solve everyday problems.
Someone at Level 0 may have little or no experience with AI. At Level 1, they're experimenting. By Level 2, they're regularly incorporating AI into their daily activities.
For most people, this is where the journey begins, and there's nothing wrong with that. The immediate benefits are substantial.
But the relationship is largely transactional. They provide an instruction and expect a result. They may not consider what happens to the information they provide, how the result was generated, or what systems are involved.
The next shift occurs when people realize that AI can do more than answer questions. It can help them create things.
These are the AI Builders, Levels 3 and 4.
People who have never written software are suddenly creating websites, applications, automated processes, and business tools that previously would have required professional developers.
I find this development fascinating. We're witnessing the removal of barriers that have existed throughout the history of computing. An idea that once required a team of programmers and a substantial budget can now become a working application in an afternoon.
But there's a consequence to making sophisticated technology easier to create.
The ability to build something doesn't automatically include the knowledge required to make it safe.
I recently encountered an example involving an attorney who used AI to create a program for publishing PDF documents. The program worked, accomplishing exactly what the attorney wanted it to do.
Unfortunately, the attorney later discovered that legal documents uploaded through the application had become publicly accessible.
Think about that for a moment.
Someone with no intention of exposing confidential information created a working application that did precisely that.
The problem wasn't necessarily that AI failed to follow instructions. The problem was that the instructions didn't account for everything surrounding the requested functionality.
Traditional software development involves more than making a program work. Developers must consider security, authentication, authorization, data storage, error handling, and countless other details that users rarely see.
AI can make those complexities less visible, but it doesn't make them disappear.
As the ability to create software becomes more accessible, understanding the consequences of what we create becomes more important, not less.
When AI Becomes Part of the System
As people gain experience, they begin looking beyond individual applications. They want AI to interact with the technology they already use.
This brings us to AI Integrators, Levels 5 and 6.
Instead of asking AI to draft an email, they want it to recognize when an email needs to be sent, determine who should receive it, and potentially send it automatically.
Instead of asking AI to analyze a report, they want it to continuously monitor the underlying data and identify problems before anyone knows they exist.
AI begins transitioning from a tool into an operational participant.
This is an important change because the consequences of mistakes become more significant.
An incorrect answer in a chatbot might be inconvenient. An incorrect action performed automatically across several connected systems could be disastrous.
Imagine an AI system responsible for managing customer accounts. It receives an instruction that it interprets incorrectly and begins modifying records across multiple applications. Each individual action might appear legitimate, but the combined result could create a substantial problem.
Suddenly, questions about permissions, accountability, verification, and oversight become essential.
And this leads to another shift in thinking.
At Level 7, AI Systems Thinkers begin looking at how everything works together.
They recognize that improving one component doesn't necessarily improve the entire system. Sometimes it introduces new problems.
Giving an AI assistant access to more information might make it more useful, but should that assistant have access to everything? Should it be allowed to make decisions independently? Who is responsible when it makes a mistake? How do we determine whether the information it uses is accurate?
These questions go beyond programming. They're about how systems behave.
A systems thinker recognizes that the intelligence of an individual component matters considerably less if the surrounding system cannot be trusted.
This is where I believe our understanding of AI needs to expand dramatically.
We're spending enormous amounts of money and effort making artificial intelligence more capable. But increasing intelligence doesn't automatically produce reliable systems.
In fact, greater capabilities can amplify the consequences of poor system design.
Thinking Beyond the Intelligence
At Level 8, AI Infrastructure Thinkers begin asking a different question.
Rather than concentrating primarily on making AI more intelligent, they ask what foundations must exist for an AI-powered world to function.
This is an area that has occupied a great deal of my thinking.
Consider something as simple as ownership.
You create a document and allow your employer to use it. Your employer allows an AI system to analyze that document, and the AI generates new information based on its contents.
Who owns the resulting information?
What happens if you leave the company? Can your employer continue using it? Can you take it with you? What happens if another AI system needs access to it?
And who gets to make those decisions?
These questions might seem abstract until you consider how much information we create and share throughout our lives.
Our financial records, medical histories, employment information, communications, photographs, intellectual property, and personal experiences all exist somewhere in digital form.
Now imagine AI systems operating across that information, combining it, interpreting it, and generating new information from it.
Who owns what? Who has permission to access it? How do we know where information originated? How do we prevent an AI system from using information it was never authorized to access?
These are not questions that can be answered simply by making AI smarter.
They require infrastructure.
Identity, ownership, permissions, provenance, security, accountability, and trust must become fundamental components of an AI-powered environment.
I increasingly believe that artificial intelligence should operate within a foundation of governed information rather than becoming the foundation itself.
The distinction matters.
If AI is the foundation, we risk building a world where the rules governing our information are determined by whichever AI system happens to be processing it.
If ownership, identity, and permissions are the foundation, AI becomes a powerful participant operating within established boundaries.
The technology can evolve without requiring us to surrender control over the information that makes it useful.
From Individual Systems to an AI Ecosystem
Once we begin thinking about infrastructure, another question naturally follows.
What happens when all these independent systems begin interacting?
This is the territory of Level 9, AI Ecosystem Thinkers.
Imagine a future where you have a personal AI assistant that understands your preferences, responsibilities, finances, health information, and professional relationships.
Your employer operates its own AI systems. Your bank has AI agents. Your physician uses AI to help manage your healthcare. Your insurance company has AI systems evaluating risk.
Now imagine these systems communicating directly with one another.
Your personal assistant negotiates with your insurance company's AI. Your physician's AI requests medical information from another healthcare provider. Your employer's AI coordinates with a supplier's AI to resolve a business problem.
None of these interactions necessarily requires a human to manually transfer information between systems.
The potential benefits are enormous.
But so are the questions.
How does one AI system establish that another can be trusted? How is permission granted? Can permission be revoked? What happens when organizations operate under different laws, policies, or economic incentives?
And what happens when two systems disagree about who owns information or what someone is authorized to do with it?
At this level, we're no longer designing individual applications or even individual infrastructure platforms.
We're considering how entire ecosystems of independently operated systems might coexist.
This requires thinking about interoperability, governance, economic incentives, competition, cooperation, and trust across organizational boundaries.
I suspect this will become one of the defining challenges of the AI era.
The technology needed to create intelligent agents is advancing rapidly. The infrastructure and agreements required for those agents to interact safely across organizational boundaries are much less mature.
We may soon have systems capable of performing extraordinary tasks without having established the rules that allow them to work together responsibly.
The Bigger Picture
Eventually, the conversation moves beyond the technology itself.
At Level 10, AI Societal Thinkers consider how artificial intelligence might change the structure of civilization.
What happens to employment when intelligence becomes increasingly available as a service? How does education change when every student has access to an AI tutor capable of explaining virtually any subject?
What happens to economic opportunity when individuals can accomplish work that previously required entire organizations?
Will AI concentrate power among a relatively small number of companies, or will it distribute capabilities that were once available only to the wealthy and powerful?
How do governments and institutions adapt to a world where intelligent systems can make decisions, negotiate agreements, and act independently?
And perhaps most importantly, how do we preserve human individuality, autonomy, and accountability in a world increasingly influenced by artificial intelligence?
These questions fascinate me because they reach far beyond computers and software.
They involve economics, philosophy, governance, human behavior, and the fundamental structure of society.
But I don't believe Level 10 represents some final destination where someone has mastered artificial intelligence.
A philosopher might recognize the societal implications of AI without understanding the technology that makes them possible. An infrastructure architect might understand the underlying systems in extraordinary detail without spending much time considering their broader social consequences.
The levels describe an expanding scope of consideration, not a hierarchy of human intelligence or expertise.
We need people thinking at every level.
The challenge is recognizing when our understanding of a particular level is insufficient to appreciate the consequences of what we're doing.
Why I Believe This Matters Now
For most of my life, using sophisticated technology required learning at least something about how that technology worked.
Software development required programming knowledge. Building business systems required technical expertise. Automating processes required understanding the underlying applications.
Artificial intelligence is rapidly removing many of those barriers.
That's an extraordinary opportunity, and I believe it will create possibilities we can barely imagine today.
But we're also entering a period where people can create and operate increasingly sophisticated systems without understanding their underlying architecture.
We're separating the ability to use technology from the knowledge traditionally required to create it.
I don't believe the answer is to discourage people from using AI or building things with it. Quite the opposite.
I want more people to experiment, create, and discover what this technology can accomplish.
But I also want them to recognize that every action can have consequences beyond the immediate result.
The attorney who created the PDF application wasn't trying to expose confidential information. The attorney was trying to solve a problem.
That's what makes the example so important.
Good intentions and successful execution don't guarantee a safe outcome.
As AI becomes more capable, the distance between what we ask it to do and the consequences of what it actually does may become increasingly difficult to recognize.
Understanding that distinction is going to matter.
Where Do You Fall on the Spectrum?
Think about the last conversation you had about artificial intelligence.
Were you talking about something AI could do for you? Something you wanted to build? How to connect it to your existing systems?
Or were you considering the infrastructure needed to support it, how independent AI systems might interact, or what all of this could mean for society?
There's no wrong answer, and your answer will probably change depending on the subject.
The purpose of the AI Vision Spectrum isn't to assign everyone a permanent ranking.
It's to encourage people to recognize that there are other ways of thinking about the same technology.
Someone who spends most of their time at Level 2 doesn't need to become an infrastructure architect. But understanding that Levels 7, 8, and 9 exist might change how they approach a decision involving their personal information.
Someone building an application at Level 4 might benefit enormously from asking a few Level 7 questions before deploying it.
And someone contemplating the societal implications of AI might develop a more informed perspective by understanding the infrastructure that makes those possibilities real.
That's the value I see in the framework.
It encourages us to look beyond the immediate task and consider the larger system we're participating in.
I believe artificial intelligence will become one of the most consequential developments in human history. But I don't believe understanding its future requires everyone to become a computer scientist.
It requires curiosity, a willingness to ask questions beyond the immediate objective, and an appreciation for the consequences of our decisions.
My hope is that more people begin looking beyond what AI can do for them today and start considering what kind of world we're collectively building for tomorrow.
Because the future of artificial intelligence shouldn't be understood only by the people building it.
It should be understood by the people who will live with it.
Continue reading: A place for individual thinking.