With the U.S. government struggling to compile data on how artificial intelligence has affected the workforce, Anthropic has taken matters into its own hands. The AI startup has released the fifth of its “Economic Index” reports, gathering data on 1 million user conversations from both its Claude web offering, Claude.ai, and its developer API. Needless to say, the findings don’t exactly paint a rosy picture for tech workers’ career prospects.
Key Takeaways From Anthropic’s Economic Index Report
In its latest Economic Index report, Anthropic found that more white-collar workers are using Claude, sales and trading tasks are increasingly automated and experienced users enjoy greater success at automating tasks with Claude than newer users while developing more collaborative relationships with Claude. These trends suggest that automation is slowly but steadily spreading, putting more pressure on tech workers to adopt AI tools to stay employable.
Among the report’s most eye-catching revelations is the fact that more experienced Claude users have greater career success than newer users, hinting at a growing divide between Claude-savvy power users and beginners. This shift serves as a warning for workers who have yet to adopt AI tools and organizations that have yet to properly prepare their employees for AI, making it all the more urgent that tech leaders develop an AI-ready workforce — or else risk intense backlash against the technology.
What Does Anthropic’s Economic Index Report Say?
Anthropic discovered some telling changes in how people are using its web product and API, with several consequential trends for tech workers to note.
More White-Collar Workers Are Using Claude
Claude is now used for a wider variety of tasks, as revealed by calculating the “average value of tasks,” or the average hourly wage of the U.S. workers performing those tasks. Between January 2025 and February 2026, the average hourly wages of Claude.ai users dipped from $49.30 to $47.90, which is still well above the overall average hourly wage of $37.30. In other words, college-educated workers remain the bulk of Anthropic’s user base, and a greater portion of the white-collar workforce is converting to Claude products.
While Anthropic attributes the decrease in the average value of tasks to new users asking questions about casual topics like sports and the weather, it notes that coders and developers are pivoting from its web product to its API as well. In fact, the number of “computer and mathematical tasks” jumped by 14 percent on the API platform and dropped by 18 percent on Claude.ai between August 2025 and February 2026.
The problem is that migrating from the web product to the API could usher in “more imminent transformation of work,” according to Anthropic. That’s because API workflows often follow more rigid rules than on Claude.ai, reducing the need for human guidance to finish tasks. And if this migration accelerates, certain professions could be acutely exposed to the effects of AI-driven automation.
More Roles Are Being Automated
Customer service workflows remain a mainstay in Anthropic’s API data, especially “automated support for payment and billing issues.” As a result, customer service representatives are most likely to be affected as AI becomes more widespread. They could soon be joined by other professions, though, as two main job categories saw their number of API workflows double between November 2025 and February 2026:
- Business sales and outreach automation: This includes researching and qualifying leads, updating customer data, drafting emails and creating sales content.
- Automated trading and market operations: This includes monitoring markets, recommending investments and sharing information on market conditions.
If tasks in these categories continue to be delegated to the API platform, business representatives, sales representatives, traders, market researchers and other relevant roles could feel the brunt of the next wave of automation alongside their customer service counterparts. The question then is how exactly AI could impact their jobs.
Higher Tenure Users Collaborate More With Claude
To understand how AI usage could reshape the nature of work, Anthropic breaks down its users into “high tenure” and “low tenure” categories. High-tenure users signed up for Claude at least six months before the survey, while low-tenure users are newer sign-ups.
The biggest difference between these two groups is how they use Anthropic’s platforms. About 49 percent of high-tenure users turn to Claude for work-related tasks, compared to roughly 42 percent of low-tenure users. Taking a closer look, high-tenure users execute “directive” tasks on Claude 29.4 percent of the time compared to 38.1 percent for low-tenure users. They also perform “task iteration” with Claude 28.2 percent of the time, compared to 24.5 percent among low-tenure users.
In short, high-tenure users are less likely to delegate rules-heavy tasks to Claude that can be easily automated. Instead, they more often use Claude to improve their work — not to do all the work for them. This behavior suggests that experienced users go back and forth with Claude, treating it as a collaborator rather than just a simple tool. High-tenure users are then finding ways to augment their abilities with AI, which might be the reason why they seem to have a far more positive experience than low-tenure users.
Higher Tenure Users Experience More Success With Claude
Anthropic introduced a variable in its previous report called “task success,” which gauges whether users can successfully automate tasks with Claude by looking at both whether a task can be automated and, if so, how many times it takes to automate it. Successful users should then be able to automate tasks with Claude in a limited number of attempts, making automation practical and cost-efficient.
Applying this variable to high-tenure and low-tenure users, Anthropic found that high-tenure users are 5 percent more likely to have a successful interaction with Claude. The company then factored in differences in the models used, languages of communication, use cases and countries where users signed on. None of these variables reduced the gap in success between high- and low-tenure users by more than a couple of percentage points, confirming that the greater success rate of high-tenure users is no fluke.
There’s no denying that more experienced users better understand how to use Claude and get the outputs they want. Tech workers who have put off adopting Claude and other AI tools may then want to reconsider their approach to avoid getting left behind as enterprises integrate the technology into their operations.
What Does This Mean for Tech Workers?
Anthropic’s report makes it clear that the tech workforce is splintering into two classes: Those who are skilled at using AI and those who are not. Experienced users also know how to enhance their work using Claude, suggesting that the immediate future of AI points to augmented intelligence rather than full automation.
That doesn’t mean that all tech workers are safe, though. Companies currently experimenting with AI agents — advanced systems that can complete complex tasks without human assistance — are already trimming their headcounts in favor of leaner teams, where a small number of employees manage networks of agents. Going forward, tech leaders will likely covet employees who know how to work with these systems, while laying off those who can’t or refuse to adapt.
Companies are also adopting AI usage as a performance metric, factoring in how often employees use AI for their performance reviews. This shift has culminated in some organizations embracing “tokenmaxxing,” or the practice of using the maximum amount of tokens possible with an AI tool. As AI usage becomes mandatory, tech workers may be left with no choice but to lean in to stay employable.
Besides learning AI skills like writing prompts and working with agents, tech workers can cultivate soft skills like communication and problem-solving, too. AI certification programs and machine learning bootcamps can also help professionals round out their hard and soft skills to prepare them for an augmented workforce. But these efforts don’t have to fall entirely on workers, with tech companies and the federal government set to play a major role in training the next generation for the age of AI.
How Can the AI Industry Support Tech Workers?
Tech companies need workers who can build and train AI models, maintain their infrastructure and supervise autonomous agents. They’d do well then to take proactive measures to prevent large sections of the workforce from being left behind in the wake of AI-driven automation. For example, Google hosted its inaugural “AI for the Economy Forum” in Washington, D.C., where it committed to three main initiatives:
- Fostering AI literacy among rural healthcare workers in partnership with the Johnson & Johnson Foundation.
- Backing a Jobs for the Future program that works with 100 companies to develop new apprenticeship opportunities.
- Partnering with the Manufacturing Institute to train 40,000 current and aspiring manufacturing workers in AI skills and offer apprenticeships in 15 U.S. regions.
A host of tech companies also teamed up with the Trump administration to devise ways to incorporate AI into K-12 education, given that American educators have struggled to keep up with technological advancement. At the higher-education level, edtech companies TED, Khan Academy and ETS have joined forces to establish the Khan TED Institute, which aims to reorient college education around AI skills training and other skills relevant to the job market — all while keeping costs under $10,000.
Still, apprenticeships, skill trainings and education programs may not be enough to convince workers that tech companies are on their side. Amid these ongoing initiatives, tensions over AI taking jobs have reached a breaking point, leading to heightened levels of anxiety, anti-AI sentiments and even acts of violence. Meanwhile, the AFL-CIO, America’s largest union federation, believes that the U.S. workforce can never truly be shielded from the worst effects of AI unless they are guaranteed stronger worker protections.
Tech companies’ current initiatives are a solid start, but they need to do more to win support for their AI agenda among workers and the general public. After all, members across the political spectrum are beginning to turn on the technology, generating momentum for data center moratoriums and tougher tech laws. If tech leaders can’t convince everyday Americans that the future of AI is designed to work for them and their career prospects, they may face an even fiercer wave of AI resistance in the workplace and across U.S. society.
Frequently Asked Questions
Which professions are most exposed to AI right now?
According to Anthropic’s report, customer service workflows are being heavily automated on Anthropic’s API platform. The biggest increases, though, occurred in business sales and automated trading, where API workflows doubled between November 2025 and February 2026. This means that roles like business representatives, sales representatives and market researchers could be exposed to greater automation in the near future.
Why are experienced users more successful with Claude?
Anthropic defines “task success” as whether users can successfully automate a task with Claude, measuring whether a task can be automated at all and how many times it takes to automate it. Notably, the types of models used, languages of communication, use cases and countries where users signed on didn’t alter the results.
A key difference, though, is that more experienced users treated Claude as a collaborator to improve upon their work, while newcomers used Claude to do the work for them. Claude power users may simply have a better understanding of how Anthropic’s platforms work and how to get the outcome they’re looking for — without relying on full automation.
What are tech companies doing to prepare workers for AI?
Google held its inaugural “AI for the Economy Forum,” pledging to contribute to initiatives involving AI literacy, apprenticeships and upskilling for manufacturing workers. Meanwhile, edtech companies TED, Khan Academy and ETS have formed their own institute to recenter higher education around AI training and job market-related skills. Various tech companies have also committed to President Donald Trump’s plan to integrate AI skills training into the American K-12 education system, preparing the next generation of AI-ready workers.
