The economic potential of generative AI: bigger than most people realize

Ask someone to guess how much value generative AI could add to the global economy every year, and you’ll get a shrug, or a wild guess, or “a lot, probably.” McKinsey actually ran the numbers. Somewhere between 2.6 trillion and 4.4 trillion dollars annually. The high end of that is bigger than the entire GDP of the United Kingdom.

I had to read that stat three times before it actually landed for me. Big numbers like that tend to slide right past you unless you stop and sit with them for a second. So here’s me sitting with it, and pulling apart where that money is supposed to come from. Because the economic potential of generative AI doesn’t spread evenly. It piles up hard in a few specific places, and those places say a lot about what’s actually shifting under our feet right now.

McKinsey didn’t just guess at 2.6 to 4.4 trillion. They built it from 63 separate use cases across 16 business functions, things like AI drafting marketing copy, handling customer service chats, writing first drafts of code, speeding up R&D work that used to take months. Four of those functions carry about 75% of the total value on their own: customer operations, marketing and sales, software engineering, and R&D. If your job lives in one of those four buckets, this stops being an abstract McKinsey chart. It’s just describing your Tuesday.

There’s a second, bigger number buried under the first one, and I think it gets less attention than it should. Once generative AI stops being a separate app you open and starts getting baked directly into the software people already use all day, McKinsey’s estimate roughly doubles. Somewhere between 6.1 and 7.9 trillion a year. That’s the real gap here, not “will companies use AI” but “will it be a tool you choose to open, or something quietly running underneath everything you already do.”

Generative AI’s economic value doesn’t land the same way in every industry

Now, the averages hide where the actual money sits, which bugs me a little because averages always do this. Banking is looking at 200 to 340 billion dollars a year, mostly from automating the risk reporting and compliance paperwork nobody was excited about doing by hand anyway. Retail and consumer goods sit at 400 to 660 billion, driven mostly by AI-written marketing content and faster customer service. Banking, high tech, and life sciences see the biggest gains relative to their revenue, and that tracks once you think about it. These are knowledge-heavy industries, and knowledge work is exactly the lane generative AI is strongest in. Put it on a warehouse floor and it’s not doing much. Put it in a call center or a research lab and it’s suddenly indispensable.

What generative AI’s economic impact actually does to a workday

Here’s the part I go back and forth on. McKinsey estimates that current generative AI tools could theoretically automate work activities that eat up 60% to 70% of the time employees spend working. Back in 2017 their estimate was around half. That’s a big jump for a few years, and it’s mostly because generative AI is unusually good with language, and language-heavy tasks eat up roughly a quarter of all work time across the whole economy.

I want to be careful here, because “theoretically automatable” and “your job is gone” are not the same sentence, and a lot of coverage treats them like they are. What’s actually true is quieter and, honestly, a bit sneakier. Drafting an email. Writing a rough first pass of code. Answering a routine customer question. These used to eat an hour. Now they take a few minutes. Multiply that across a whole company and the freed-up time has to land somewhere, whether that’s more output, fewer people, or, in the best case, people spending time on the stuff that actually needed a human in the first place.

On plain productivity, generative AI alone could add 0.1% to 0.6% to labor productivity growth every year through 2040. Stack it with other automating tech and that climbs to an extra 0.2 to 3.3 percentage points annually. Small numbers, I know. But compound anything over fifteen years and it stops looking small.

The asterisk nobody puts in the headline

And here’s my actual gripe with how this story usually gets told. Every version of these stats carries a quiet condition: “if managed well.” Nobody puts that in the headline. The economic potential of generative AI doesn’t show up automatically just because the technology exists on a server somewhere. It shows up when companies actually rebuild how work flows instead of dropping a chatbot into a process that was broken to begin with, and when workers get real support while their roles shift instead of getting handed a login and a shrug.

That’s the detail most coverage skips entirely. The trillion-dollar figures are a ceiling, not a promise. Getting anywhere close to that number takes real money spent on retraining people, real time spent redesigning workflows, and a kind of organizational patience that, if I’m honest, most companies I’ve read about do not have.

So what do you actually do with this

If you run a business, don’t try to overhaul everything at once. Start with the parts of your operation that already look like one of those 63 use cases and work outward from there. If you’re building a career, take a hard look at your week. Which parts look like drafting, summarizing, or answering the same three questions on repeat? And which parts depend on judgment, or trust, or a relationship you’ve built with a client over years? Protect that second category. Sharpen it. It’s the part nothing in this whole report can do for you yet.

The economic potential of generative AI isn’t hype, at least not the way the term usually gets thrown around. It’s a specific, sourced number tied to specific tasks in specific industries. What happens with it, whether any of that value actually lands in your paycheck or just your company’s margins, has less to do with the technology itself and a lot more to do with what gets done with it over the next year or two.