We just released our third bi-annual AI Proficiency Report, the only report that measures AI proficiency (not just usage) in the knowledge workforce.

What we found: AI adoption is surging, but most people are still very bad at using it. They don’t know how to prompt, they don’t understand how AI works, and they aren’t going beyond basic use cases.

This is going to be a huge problem for organizations expecting ROI from their AI investments. 80% “adoption” won’t matter if everyone is using AI like Google search.

We need to start looking beyond usage (especially self-reported usage) and examine how people are actually using these tools. Workers need help to figure out AI, and right now companies aren’t providing it.

Here’s my take on our newest data, and what you should do as an individual and a leader.

Adoption is accelerating

Companies are moving faster in terms of AI adoption. Maybe they’re listening to Tobi Lutke or defending their AI strategy to the board, but CEOs have upped their approval and investment in AI over the last 6 months.

• Company approval of AI is up 12% in the last 6 months
• LLM deployments are up 5%
• Employees receiving AI training are up 80%

And employees are more excited and more active too. Employee excitement about AI is up 2x, and 55% of knowledge workers are using it at least weekly.

This confirms what I’ve been thinking for a while – we’re leaving the “early adopter” stage of AI and entering the “early majority.” In other words, AI is going mainstream. You won’t seem like a geek for saying you use AI every day, as evidenced by ChatGPT’s nearly 1 billion users.

But all this adoption masks a big problem.

AI proficiency is still very low

To measure AI proficiency, we put knowledge workers through an actual test: We ask them to give sample prompts for real-world scenarios, and we ask them about the complexity and depth of their AI use cases.

This gives us a more accurate read on what the workforce is actually doing with AI, vs. just asking “do you use it?”

Our newest data says: Yes, people are using it. But they’re not using it well.

Based on our research, only 10% of the workforce is “AI-proficient.” The rest are essentially beginners – people with poor prompting skills who don’t understand how AI works or use it regularly.

Most of their prompts are also really bad. People are treating AI like it’s Google or Siri. You’re not going to get any value with prompts like the ones shown here.

And maybe most concerning for enterprise, most people misunderstand basic AI concepts. 44% don’t know what a hallucination is (crazy to me, given I feel like I read about the problems with hallucinations constantly).

Some other stats that stuck out to me in the research:

• 25% of employees still don’t know what to use AI for – beyond generating some copy or summarizing meeting notes.
• Every industry, even big tech, is behind. The average proficiency score for B2B and consumer tech is 47/100 – an F by any measure.
• Functions that will be disrupted by AI soon aren’t using it well. Marketing scores a 41/100, and customer support is dead last with a 33.

Maybe not surprising, but concerning, is most people overestimate their AI skills considerably. 54% of people think they’re proficient – only 10% are.

Modest gains, but lots of upside

We’re seeing some gains from AI, but only the beginnings. 34% of workers say they save 4+ hours a week with AI (though take that self-reported data with a grain of salt).

But they’re doing it with mostly basic use cases. 62% of workers use AI as an assistant. Only 30% use it as a thought partner or a creator.

And even AI experts aren’t tapping into AI’s full potential. Less than half of experts are using advanced tools like Deep Research and custom GPTs, which should be table stakes for proficient users.

My take on this: People can’t keep up. OpenAI has released seven new models in the last few months alone – because they’re prioritizing investor attention over what users actually need.

ChatGPT’s o3 model gets consistently better outputs, but only ⅓ of users use it. People are trying, but it’s hard to stay on top of it when there’s a new release every other week.

You can’t grow AI proficiency organically

If you’re a company leader, this is the biggest takeaway: AI proficiency must be engineered, and most companies are doing a bad job at it. Here’s what the data says you should be doing.

1. You need a clear AI policy.

AI experts are 3x more likely to have a company AI policy than AI skeptics. You should have a data policy AND a manifesto – download our manifesto creation guide if you don’t know how to write one.

2. Your managers need to be bought in and encouraging AI use.

If the CEO is saying “use AI” but managers are saying “don’t worry about it,” you have a problem. The vast majority of AI skeptics say their manager is silent on AI, or prohibits its use completely.

Make sure your managers are bought into your vision on AI, and do regular surveys to understand if they’re encouraging use at the ground level.

3. LLM access matters.

80% of AI experts have access to a paid company LLM, compared to 28% of AI skeptics. Give your people access to ChatGPT or Claude Enterprise (the best tools, in my opinion) and make sure everyone gets them. We see that access is often given to senior employees, with ICs left to figure it out for themselves.

‍4. Training matters.

We’re not just biased because we do AI upskilling: training matters. 71% of experts have access to company AI training, compared to just 10% of skeptics. Give your people access to training that goes beyond “checking the box” on compliance and helps them surface real use cases they can put to work.

Your mandate as a leader

It pays to be an AI expert. 84% of AI experts save more than 4 hours a week – only 24% of novices do. So that starts with you.

Your job is not to practice “performative AI” – putting out press releases, mentioning it on earnings calls or all-hands meetings, etc. Your job is to create as MANY internal AI experts as fast as possible.

And if you’re an individual contributor, your job is to BECOME an expert. This is the only way to get real ROI from AI, either as a worker or an organization.

The steps to do it are simple, but they take time, commitment, and repeating your vision and commitment about 1,000 times (trust me, I’ve led my own team through AI transformation).

To get started, download our report. If you’re a leader, a manager, or in any way accountable for an AI deployment, you need to look at this data firsthand. There are clear trends here that are too costly to ignore.

Then make a plan – starting with your AI manifesto, if you don’t have one. You can do this. And if you need help, reach out – greg@sectionai.com.

P.S. we can run the same benchmark for your team of 50+ people. You’ll not only get a pulse on your own team’s AI proficiency, you’ll get to benchmark your efforts against the rest of the workforce (and your competitors) and use the data to set a strong strategy. Reach out to learn more.