Where the money comes from today
AI is already a big business. Anthropic reported a $65B annualized run-rate in July (CNBC), and OpenAI was at about $50B in late September (Calcalist). Together that is about $115B a year. That is a run-rate, not booked revenue, and the two companies count revenue differently, so treat the total as a ballpark.
Software work is where it took hold first. In Anthropic's Economic Index, computer and mathematical tasks were the largest single category of Claude conversations in May 2026, at 23.8%. No other job category came close.
The people behind that work are a small group. SlashData counts 47.2 million developers worldwide, and about 21 million of them are professionals.
The next wave is everyone else
Gartner estimated there would be 1.14 billion knowledge workers worldwide by 2023. That is more than 20 times the number of developers. Most of them write proposals, read contracts, quote jobs, answer customers and run the back office. They don't write code.
That work is already showing up. In the same Economic Index data, tasks from sales (9.14%), office and administrative support (7.89%), management (5.9%) and business and financial operations (5.77%) together made up 28.7% of Claude conversations.
Big firms are paying to get there first. According to the Financial Times, Kirkland & Ellis said it will spend $500 million over three to four years to build its own AI technology stack.
Why it hasn't reached your firm yet
In the Census Bureau's Business Trends and Outlook Survey, 23.8% of US firms said they used AI in the past two weeks (September 7 to 20, 2026). Among firms with 250 or more employees it was 44%. Plenty of firms have tried it. Very few have rebuilt how they work around it.
The gap is implementation. The model companies sell the models. Nobody at those companies is coming to sit with your estimators or your claims team, wire AI into your systems, or teach your people to trust it.
What this means for owner-led firms
If you run a firm of 50 to 500 people, here is how I read all this.
The wave is aimed at you. Developers went first because the tools were built by developers, for developers. The next billion users do the kind of work your people do every day.
You won't be early, but you're not late. Most US firms didn't use AI in that two-week window, and most that did are still dabbling. Between dabbling and running differently there is still a wide gap, and it's open.
Implementation is the bottleneck. The models are good enough for a lot of this work now. What's missing is the hands-on work of wiring AI into your systems, your rules and your people. Kirkland can spend $500 million on that. You can't, and you shouldn't need to.
Put money into growth as well as savings. For an owner, that means better proposals, faster quotes and more capacity per person, along with any savings from doing the same work with fewer people.
The takeaway
AI's first big wave was software. The next is more than a billion people who do work like your team's, and it is just getting started. The firms that win won't be the ones with the fanciest tools. They'll be the ones that pick one or two real workflows this year, put someone close to the work, and make AI part of how the firm actually runs.
That is the work Setanta does. If you want to talk about where to start, I'm easy to reach.
Gary O'Neill