What Anthropic's AI Economic Study Means for Your Pay
11 Sept 2026
Half of all knowledge work. Done by AI. By 2030.
Half of all knowledge work. Done by AI. By 2030.
That is close to what the average American expects.
A much awaited study by Anthropic just came out the day before yesterday. (scroll down for link)
Anthropic asked more than 10,000 people in the US in August, then built the answers into a model of the economy running to 2030. Three futures came out of it.
→ Modest: AI arrives the way the internet did. Wages and unemployment stay in normal ranges.
→ Substantial: AI does half of all knowledge work. The economy grows at twice its normal rate. Unemployment sits around 5 percent.
→ Extreme: AI passes humans on most knowledge tasks. Unemployment spikes past recession levels and knowledge worker wages fall by more than 10 percent by 2030.The typical survey answer was on the middle one.
And in that middle one, knowledge worker wages stay flat while everyone else gains. Nobody gets fired in that scenario. The work just stops being worth more each year.
I do not read this model for the GDP number. I read it for that wage line.
My clients are knowledge workers. Directors, chiefs, heads of function. Every one of them has been told to adopt AI. Almost none of them have been told what happens to the price of what they know. The model runs on averages.
It does not track individual people, so it cannot tell you who ends up flat and who does not. My bet is on the ones who use these tools in everything they do and still keep their own judgement. Thinking for themselves. Reading the room.
Which future do you expect?
PS the model is public and you can run your own assumptions through it:
https://www.anthropic.com/institute/econ-scenarios
PPS: I write about this in my newsletter, sign up here:
https://www.thechangerepublic.com/#subscribe
How this shows up in the coaching work I do with senior leaders
The scenario I keep rereading is the middle one, the one most people in Anthropic’s survey expect. Nobody loses their job in it. The economy grows. And knowledge worker pay simply stops climbing while every other kind of work gets a raise. It is the least dramatic outcome in the study and, for my clients, the most likely one to arrive quietly and go unnoticed until a few years in.
My clients are exactly the group this model is describing: directors, chiefs, heads of function. Almost every one of them has been told to adopt AI this year. Almost none of them have been told what adoption does to the market value of their own judgment. The model runs on averages, so it cannot tell you which of the group ends up flat and which one does not. That part is not economic, it is personal, and it is the part I work on.
The distinction that keeps showing up in my Private AI Advisory work is between using AI and outsourcing judgment to it. The leaders who are likely to sit above the average in this scenario are the ones who use AI in everything and still bring their own read of a room, their own call, their own accountability to the table. That is not a skill you keep by accident. It is one you practise, the same way you would practise any other part of the job that matters.
If this resonates and you want a practical way to start that practice yourself, I have a free AI Coach Tünde Light tool, a set of light AI-powered self-coaching prompts to help you reflect on where you are using AI well and where you might be handing over more than you meant to. Leave your email here and I will send it over: https://www.thechangerepublic.com/free-resources