Will AI Replace Knowledge Workers by 2030? Anthropic's New Forecast, Explained

Anthropic's Economics team has released a model that puts a number on something people have only been guessing about. In their middle scenario, AI is capable of doing half of all knowledge work by 2030, and the majority of it autonomously. The work sits in a technical report called Economic Scenarios for Transformative AI (Korinek et al., 2026), released alongside a public scenario explorer in September 2026.

Half of all knowledge work. That is the headline everyone will run with.

But there is a second half of that sentence, and it is the more important one. Even though AI is capable of doing that work, it is not adopted for all of it. Most knowledge work tasks in that scenario are still being done without AI.

I’m Riten, founder of Fueler, a platform where people get hired based on proof of work. I spend my days looking at what companies actually ask for when they hire. That gap between what AI can do and what AI is actually used for is the most useful thing in this entire report, and almost nobody is talking about it.

What Anthropic's Three Scenarios Say About Knowledge Work

The model does not give one prediction. It gives three, because the future depends on how fast AI improves and how fast people and companies actually pick it up.

Modest scenario. AI has roughly the same kind of impact the internet did. Real gains, but within the historical norm for new technology, and they arrive gradually. GDP ends up 1.6% higher than it would have been, at $34.1 trillion.

Substantial scenario. AI makes a bigger impact than the internet or the railroad. This is where AI is capable of half of all knowledge work by 2030. The economy grows at twice its normal rate. GDP is 8.3% higher, at $36.3 trillion. Wages for knowledge workers do not rise. Other workers see gains.

Extreme scenario. AI is more productive than humans at the vast majority of knowledge work tasks, does nearly all of them on its own, and creates essentially no new knowledge tasks for people. The report says this would likely require recursively self improving AI adopted quickly. Annual GDP growth reaches 15%, doubling the economy every 4.5 years. GDP is 32.4% higher, at $44.4 trillion. Many fewer people work in knowledge work, and unemployment rises beyond typical recessionary levels.

Anthropic also surveyed more than 10,000 Americans in August about their expectations. The typical respondent's answers landed close to the substantial scenario. Around 10% held views in line with the extreme one.

So the crowd's best guess is the middle path. Big change, not collapse.

Five Things the Forecast Actually Tells Knowledge Workers

Reading past the headline number, five things stand out that change how I would plan a career right now.

Capability is not the same as replacement. In the substantial scenario, AI can do half of knowledge work and still most of that work happens without it. Companies move slower than technology. Budgets, training, trust, compliance and habit all slow adoption down. That gap is time, and time is what you use to reposition.

Automation happens task by task, not job by job. The model treats every job as a bundle of tasks drawn from the US Department of Labor's O*NET taxonomy. Some tasks get automated, some get augmented, some stay untouched, and some new ones appear. Your job does not vanish. Four of your eleven tasks change hands.

New tasks are part of the picture in every scenario except the worst one. The report notes that historically new technologies create new tasks for workers. Reviewing an AI proposed plan, checking AI output for errors, directing AI tools. Only in the extreme scenario does the model assume essentially no new knowledge tasks are created.

Flat wages are the realistic risk, not unemployment. In the substantial scenario, wages for knowledge workers are essentially flat while other workers gain. Flat pay for five years is a quieter problem than job loss, but over a career it costs just as much.

In most scenarios, unemployment stays inside historical ranges. Job reallocation and unemployment both stay within ranges history has already seen, with one exception. Only in the extreme scenario does unemployment spike to historic levels. The default future in this model is disruption, not disaster.

Why It Matters

If you are a writer, designer, analyst, marketer, editor or developer, you have probably had a version of this thought at 2am. This report is the first serious attempt I have seen to put structure around it instead of vibes.

What it tells you is that the threat is real but the shape is different from what most people picture. You are unlikely to be told "your job no longer exists." You are much more likely to find that the tasks which used to justify your rate now take a fraction of the time, and that nobody is quite sure what you are being paid for anymore.

That is a proof problem before it is a skill problem.

Anthropic is careful about the model's limits, and I think that honesty matters. It leaves out policy responses, business cycles, financial market disruptions and hyper capable robots. It does not follow individual workers, so it can only paint a coarse picture of what displacement costs a person. Reviewers pushed back in both directions. Some said the extreme scenario reads better as a thought experiment. Others said the modest one understates what is already visible in the data.

Nobody knows exactly which path we are on. That is the point of publishing three.

What This Means for Your Career Portfolio and How You Get Hired

Here is where I have skin in the game, and where I think the practical answer sits.

If half of knowledge work becomes doable by AI, then the question every hiring manager starts asking is simple and brutal. What part of this person's output still needs this person?

A resume cannot answer that. A resume is a list of titles and dates. It was built for a world where holding a job for three years was itself the evidence. In a world where the tasks inside that job changed twice, the title tells you almost nothing.

What answers it is evidence. Three moves I would make this quarter:

Show your judgment, not just your output. AI can produce a deliverable. What it cannot produce is the record of why you chose this direction over four others, what you tested, what failed, and what you learned. When you build a proof of work portfolio step by step, the description matters as much as the artefact.

Make your AI usage visible instead of hiding it. I see people quietly using AI and leaving it out of their portfolio because they think it weakens them. It does the opposite. A person who can add their AI stack to a project and explain exactly how they directed the tools is demonstrating a skill most candidates do not have yet. Adoption lags capability, and the people who close that gap first get hired first.

Build proof faster than you build credentials. Certificates prove you attended. Projects prove you can do. If you are early in your career, the student guide to building your first portfolio is a more direct path to a first job than another course, because it produces something a hiring manager can actually inspect.

For designers specifically, the same principle shows up in how the strongest ones present themselves. Look at how designers build their portfolio and you will notice the good ones sell their thinking, not their pixels.

Final Thoughts

Will AI replace knowledge workers by 2030? Based on Anthropic's own model, mostly no. Capability outruns adoption, unemployment stays in historical range in the two more likely scenarios, and the economy gets bigger in all three.

But that is not the same as saying nothing changes. Wages for knowledge workers stay flat in the middle scenario. Tasks move around inside every job. And the people who cannot show which tasks are still theirs will find it harder to argue for their value, their rate and their next role.

I built Fueler because I think hiring should run on evidence. This forecast makes that case better than I ever could. The next few years will not sort people by job title. They will sort people by whether they can prove what they contribute.

Start with a list of your tasks. Then go build a career portfolio that actually gets jobs around the ones that are still yours. That is a good use of the gap between what AI can do and what companies are actually doing with it.

Frequently Asked Questions

Will AI replace knowledge workers by 2030?

Anthropic's model suggests full replacement is unlikely in its two more probable scenarios. In the substantial scenario, AI is capable of half of all knowledge work by 2030, but most of that work is still done without AI because adoption trails capability. Only the extreme scenario, which would likely require recursively self improving AI, sees knowledge work jobs shrink sharply.

What is the difference between AI capability and AI adoption?

Capability is what AI is technically able to do. Adoption is how much people and companies actually use it. Anthropic's model treats these as separate inputs, and the gap between them is large. That gap is why knowledge work can be technically automatable and still mostly performed by humans in 2030.

Which knowledge work jobs are most at risk from AI?

The report specifically names coders and call service centre agents as roles that may face displacement, and suggests some workers move toward occupations like electrician and nurse that are less exposed to AI. Risk depends on the mix of tasks inside a role rather than the seniority of the title.

Will AI lower salaries for knowledge workers?

In the substantial scenario, wages for knowledge workers are essentially flat while workers in other occupations see gains. In the extreme scenario, knowledge worker wages fall by more than 10% by 2030. Average wages across the whole economy rise in all three scenarios, which hides this split.

How can knowledge workers stay employable as AI improves?

Make your contribution visible at the task level. Document your process and decisions, not just your finished output, and show how you use AI inside your workflow rather than hiding it. Evidence of judgment is the part of knowledge work that automated output does not replace.

Why 100,000+ professionals use Fueler

Fueler helps professionals showcase proof of work through projects, assignments, case studies, and achievements.

• Thousands of professionals use Fueler to create their digital portfolio
• Thousands of projects are published on Fueler. Check here
• Startups and Companies hire through proof of work on Fueler
• Used by freelancers, creators, marketers, video editors, writers, designers, and product managers

Our mission is to help the next 100 million professionals build a verified professional identity through proof of work