BEHAVIOR, NOT PROFICIENCY
There's a word that keeps showing up in major AI studies recently.
It's not "proficiency." It's not "skill." It's not "competency."
It's "behavior."
KPMG and UT Austin used it. Anthropic used it. McKinsey used it.
(AI Mindset has been using it for three years. Just sayin'.)
You've probably noticed something:
Researchers keep analyzing millions of real AI interactions, looking for the holy grail of what makes a great AI user.
What you may not have seen is that they keep finding the same thing:
This is about behavior. Not how fast they climb a learning curve.
And that distinction changes everything about how we should be teaching AI.
THE FUNDAMENTAL FLAW IN AI ADOPTION
Here's the problem: Almost every organization on the planet is teaching AI as if it's another tool to learn.
And why wouldn't they? That's how everything else has worked. Forever.
Think about how you learned Excel. You started by clicking on a cell and adding things up. Then you learned formulas. Then pivot tables. Then, if you were feeling particularly adventurous, macros. There was a learning curve. You climbed it rung by rung.
Think about how you learned French. Single words. Then verbs. Then short sentences. Then longer sentences. Then you ordered a croissant in Paris and felt like a god.
Every tool, every language, every piece of software in history has followed this same pattern: basics first, then intermediate, then advanced.
AI does not work this way.
And that's what's tripping up the entire enterprise world.
THE EVIDENCE IS PILING UP
Let's look at what the research is actually telling us.
KPMG US partnered with researchers at the University of Texas at Austin to analyze over 1.4 million AI prompts from 2,500 employees over eight months. Their findings were published in Harvard Business Review. The headline finding: only about 5% of users were truly sophisticated. And what separated them wasn't how often they used AI. It was how they behaved. They iterated. They guided the model's thinking. They treated AI like a reasoning partner, not a search bar.
Then Anthropic published their AI Fluency Index. Note the title carefully. They didn't call it a "Proficiency Index." They called it a "Fluency Index." And they explicitly framed their findings around behaviors, not skills.
Across nearly 10,000 conversations, the number one behavioral indicator of AI fluency - at 85.7% - was iteration and refinement. Going back and forth. Those conversations showed double the rate of every other fluency behavior.
Not a single technical skill made the list. The top indicators were things like clarifying your goal, providing examples, setting the terms of the interaction. These are all things you do when you work with a smart colleague.
Both studies arrived at the same place: This is not about climbing a learning curve. It's simply fundamentally a different thing.
WHY THIS IS SO DIFFERENT FROM EVERYTHING ELSE
With every other tool in history, there was a body of knowledge to master. Buttons to learn, stuff to understand, a prerequisite chain: you couldn't do Step 5 until you'd done Steps 1 through 4. Heck, that's the entire education system.
AI has none of that.
You can go all-in instantly. Because all it is, at its core, is talking to it like a person. And you already know how to do that.
The hurdle is not in the skills. The hurdle is in the behavioral adoption. It's in getting your brain to understand why this is so fundamentally different from everything else you've ever used.
Your brain sees a text box, and your brain says: "Gotcha. This is Google." And then your brain does what it's been doing for two decades: type a short command, get a response, walk away.
So the gap isn't about what you don't know. You gotta think about the behavior gap.
And you don't close a behavior gap with a training manual.
WHY TRADITIONAL TRAINING FAILS HERE
This is why telling people to "use AI" doesn't work any more than telling someone to eat less and exercise. We know that part! It's also not about motivation - you don't think I wanna look amazing at the beach? Or make sure my heart is healthy? I want those things! And it's not about having the best tool - I have the best treadmill around!
Behavior is different. And AI is behavior.
It's also why power users can't easily transfer their ability to others. I've seen this a million times. You take your best AI user, put them in front of a room, and they show everyone what they do. And the audience nods along. And then they go back to their desks and nothing changes.
Why? Because of what we call the Curse of Knowledge - from the 1990 Stanford study. Power users can't explain why it's easy for them and hard for others - because it's not knowledge they're transferring. It's behavior. And they don't even know they changed their behavior. It just clicked for them one day.
This also means there's no such thing as "beginner, intermediate, expert" with AI. Not really. This doesn't follow a traditional learning curve. I've seen total novices become power users literally overnight.
Because something shifted in how they perceived what AI actually is, and it was an immediate unlock.
AI NEWS OF THE WEEK
1. Anthropic Proves the Behavior Argument — With Data
Anthropic’s new Economic Index Learning Curves study found that experienced AI users get a 10% higher success rate in their conversations than newcomers. The longer you’ve been using it, the stronger the effect. fluency compounds.
2. Gemini Just Got Embedded Into Everything Google
Gemini can now synthesize information from a user’s emails, files, chats, and calendar to auto-generate fully formatted documents and build complex spreadsheets from natural language prompts. This is the quiet story of the month. Behavior change just got a lot easier, or a lot more unavoidable, depending on how you look at it.
3. OpenAI Is Becoming an Enterprise Software Company
The Financial Times reports that OpenAI plans to nearly double its workforce to around 8,000 employees by the end of 2026, hiring across product, engineering, research, sales, and technical ambassadorship. The model race still matters. But so does turning AI into software that enterprises will actually buy, deploy, and renew year after year. The frontier labs are growing up.
HOW TO ACTUALLY CHANGE BEHAVIOR
So if it's not about learning, what is it about?
It's about changing what your brain sees when it looks at the screen, and engaging the part of your brain responsible for thinking and adapting and communicating, not the part that memorizes steps and procedures.
That's why we focus on what the brain is seeing, how the brain is interpreting it, and how to get people using the right cognitive mode: The mode for having a conversation with a very smart colleague.
And when you frame it that way, the muscle memory develops fast.
WHAT THIS MEANS FOR YOUR ORGANIZATION
If you're a leader trying to scale AI adoption, you're not going to find your answer in weekly active users. Anthropic just proved that's attendance, not fluency. Someone opening the tool tells you nothing about whether they're using it well.
Measure conversation depth instead. Are your people iterating? Are they going back and forth? Are they treating AI like a thought partner or like a search engine?
And rethink your training entirely. Stop treating this like a software rollout. Stop building proficiency programs. This is behavioral change. The playbook is closer to Atomic Habits than it is to a Salesforce certification.
If your organization is still treating AI as something to learn, you're solving the wrong problem.
It's not about proficiency. It's about behavior.
Let's change the approach. We got this.
