Most People "Using AI" Are Doing the Equivalent of Typing in a Spreadsheet and Calling Themselves a Data Analyst

Most people who think they're "using AI" are exactly that: typing names into a spreadsheet and calling themselves data analysts.

It's not an insult, just where most people are. And it matters, because the distance between opening a tool and actually operating it is where the gains are.

Every debate about AI adoption is really two debates running simultaneously.

The first is about usage: Have you opened ChatGPT, used an AI feature in an app, or asked a chatbot something? The second is about proficiency: Can you reliably use AI to build something, like a workflow, an automation, or a system that produces consistent output?

Usage spreads fast, and proficiency compounds slowly. Mixing them up produces either wild overconfidence ("everyone's using AI now") or dismissiveness ("it's still just a toy"). Both miss the picture.

This isn't new. Spreadsheets are on nearly every computer on earth. Most people use them. Very few build functional systems inside them, and the same goes for CRMs, automation tools, and basic coding. The pattern repeats: a tool reaches mass awareness long before mass capability follows. AI is on the same curve, just moving faster.

Microsoft's AI Economy Institute reported global adoption of generative AI tools reached 16.3% of the world's population in H2 2025, up from 15.1% in H1. Roughly 1 in 6 people have actively used a generative AI tool. That sounds like a lot until you remember that "used" can mean opened it once, asked one question, and never returned.

That number also has real blind spots. It excludes AI embedded inside software people don't recognize as AI, enterprise tools behind corporate logins, local and offline models, and AI-powered features people interact with unknowingly. The actual footprint is probably larger than 16%. Most of it is invisible and passive.

On the enterprise side, McKinsey's 2025 State of AI survey found that 88% of organizations report using AI in at least one business function, up from 78% the prior year. Only 7% had fully scaled it. Two-thirds are still in experimenting or piloting mode. Adoption is fast, but that still doesn't tell you how many people inside those organizations know what they're doing with it.

Developers have the most reason to go deep. Stack Overflow's 2025 Developer Survey found 84% using or planning to use AI tools, with 51% using them daily. Even in the group with the most reason to go deep, daily use is barely past half. The general population is probably well behind that.

Stack the data properly and a clear shape emerges.

At the base is passive exposure. That's autocomplete in your email client, recommendations in your streaming app, and fraud detection running silently behind your bank transactions. Most people are inside the AI ecosystem without knowing it or choosing it. That base is enormous and growing.

One level up are active users, people who consciously open AI tools and do something with them. That's the 16% figure. That's real, but still a minority, and heavily skewed toward younger, educated, urban, English-speaking demographics.

Above that are operators, people using AI daily to produce output. That's likely 1–2% globally, which is small enough that most people in most professional rooms haven't met one.

At the top are deep builders, who design systems, write automation, and integrate AI into workflows that run without them. They're under 1% of the population. They have an advantage, and it comes mostly from starting earlier and going deeper.

The pyramid won't flip overnight, since the base keeps growing and the top will probably stay thin longer than most people expect.

AI proficiency develops in a pattern that rarely gets addressed directly.

Early gains arrive fast. A few hours in, most people see productivity bumps: faster drafts, quicker research, less time on repetitive work. That early return is real. It's also what creates overconfidence. People feel the tool working and assume they've arrived somewhere. They haven't.

Real leverage means compounding output, not just faster individual tasks. It typically appears between 300 and 700 hours of genuine, output-focused practice. That isn't passive use or occasional prompts. It's deliberate building: trying things, breaking them, iterating, and shipping.

System-level mastery, where someone can design and deploy AI-integrated workflows that run reliably at scale, shows up past 800 to 1,500 hours. At that point the tool stops feeling like a productivity aid and starts functioning like an entirely different way of working.

Most self-described power users are sitting somewhere in the first 50 to 100 hours of output-producing experience. The gap between where people think they are and where the ceiling actually sits is very wide.

Most people place themselves higher on this curve than they are, which is human nature. Here is an honest map:

Most people reading this sit somewhere between Casual and Regular. The move worth making is figuring out exactly where, because the distance between Regular and Operator isn't about intelligence or technical background. It's almost entirely about accumulated reps.

If you're actively building with AI right now, you are not competing with most people. Most of them are in the passive base of the pyramid.

The people you're actually up against are builders, operators, and system integrators. That's a much smaller pool, and it's easier to stand out in.

This moment is a lot like 2007 for the iPhone. It's not experimental and it's not just researchers and hobbyists. But it's early enough that most people haven't built anything real with it, the playbooks aren't written, and the people putting in serious reps now are building an advantage that will be hard to copy in three years, when the tools are easier and awareness has caught up.

The iPhone analogy holds because of what happened after 2007. Knowing the iPhone existed didn't make anyone money. People who built on it early, with apps, workflows, and businesses designed around mobile, got advantages that lasted for years. Awareness didn't set anyone apart, and building something did.

Usage will level out, and it's already happening. Within a few years, "using AI" will mean about as much as "using a spreadsheet". Everyone will assume it.

What won't normalize at the same rate: the compounded experience of people who started building early, developed real workflow depth, and put in the hours before the tools were easy and the playbooks existed.

The spreadsheet comparison holds one more time. The people who built real systems in Excel in the 1990s didn't lose their advantage when Excel became ubiquitous. They deepened it, and they became who organizations called when things needed to work. The tool spreading didn't close the gap. It widened it, because everyone had the same tool and the difference was down to skill.

Right now, proficiency still sets people apart. That will fade as the tools get easier. The operators who built early will have things that are hard to copy later: compounded reps, their own process knowledge, and proof of output.

The question isn't whether AI is worth learning, because that one is settled. The question is which rung you're on, and whether you're moving.