Anthropic Says These Jobs Are Most at Risk from AI

In March 2026, Anthropic released research into the labor market impacts of AI. In the report, the company listed the jobs that it believes will be most at risk of AI-related displacement.

The report uses the metric “observed exposure” to examine which tasks:

• Could theoretically be done by LLMs
• Are actually being automated right now

Anthropic’s report listed 10 occupations that it defined as having the most observed exposure: 1

Anthropic’s observations largely match the reality many white-collar workers face in 2026. Companies are increasingly looking to implement AI into their workflows, and there’s mounting evidence that the jobs high up in their list are relying on AI.

Computer programmers

In the 2010s, many Silicon Valley coders mocked those worried about losing their jobs to automation. “Learn to code” became the go-to response to the concerns of apprehensive workers (often those in blue-collar professions, such as manufacturing).

Now, the tables have turned.

In a 2025 survey from Stack Overflow of software developers, 84% of respondents said they were using or planning to use AI tools in their development process. Over 50% of professional developers said they used AI tools daily.

While some developers are eager to outsource coding work to AI, others are worried that it’s the beginning of the end. According to ZipRecruiter, the average annual salary for a computer programmer in the US is $64,000. 2 If an AI coding tool like Anthropic’s Claude Code can do the work of a team of developers worth thousands of dollars, who will a company realistically pick?

Amy Surrett, an engineer in Greenville, South Carolina, told Business Insider that Claude Code completed a task that would’ve taken her 2–3 days within just over an hour.

“It felt like the point of no return,” she said. “This industry is not going to be the same. My job is not going to be the same.”

On the anonymous professional networking and discussion app Blind, developers are panicking, as tech companies continue laying off thousands of employees. Sunguk Moon, a founder and the chief executive of Blind, told The New York Times that “the whole mood has changed.”

“It went from personal career planning to mass anxiety. To users talking about how hard it is to stay motivated when they might lose their job very soon, maybe tomorrow.”

Customer service representatives

Human-facing roles in customer service are also at a high risk of displacement. AI tools, particularly chatbots, are increasingly handling customer inquiries on behalf of companies.

Verizon CEO Dan Schulman recently said in an interview with Bloomberg that AI would displace “a large percentage” of customer service jobs. Schulman stated that the company has been experimenting with AI customer service representatives. He claimed that Verizon’s customer satisfaction rate is now “1,280 basis points better” than it was before.

Schulman’s beliefs aren’t unusual. A recent report from global market research and advisory firm Forrester predicted that AI will cause almost half (49%) of current customer service jobs to disappear by 2030.

Data entry keyers

If one job on Anthropic’s list seems most appropriate for AI displacement, it’s data entry. For years, it was the perfect entry-level job, with plenty of opportunities for remote work.

Now, the number of data entry gigs has decreased dramatically, and this trend seems to be continuing. In 2024, the US Bureau of Labor Statistics estimated that the number of data entry jobs would decrease by 25% by 2034. 3 Last year, the CEO of the software company SAP said manual data entry would disappear from the company by 2027.

Moreover, many firms that traditionally hired human data entry keyers have pivoted toward AI. For instance, the Canadian technology company TELUS Digital, which used to hire freelancers for manual data entry gigs, now focuses on data annotation (i.e., labelling objects in images, words in bodies of text, etc.) and AI training.

While many of the jobs mentioned in Anthropic’s list are being outsourced to AI, there is also evidence to suggest that technology can’t entirely replace humans—at least, not yet.

For many businesses, working with AI is proving more difficult than they originally thought.

The unexpected cost of AI implementation

One of the biggest purported benefits of AI tools for business is that they reduce costs. Companies will theoretically be able to consolidate their workforce and let AI handle their tasks.

Many businesses, such as Microsoft and Uber, encouraged their programmers to use Claude Code as much as possible earlier this year.

However, some of those companies have had to scale back their employees’ AI use after they realized just how much the technology was costing them.

Microsoft recently cancelled the majority of its Claude Code licenses only a few months after it told workers to embrace the technology, and Uber burned through its entire 2026 AI budget in just four months. In some cases, it turns out that AI is actually more expensive than human workers.

The problem is that to use AI tools like Claude Code, you need to spend “tokens.” Each action costs a set number of tokens, and the more complicated a task is, the more are needed.

Currently, AI’s potential to drastically cut down labor costs in software development seems to still be hypothetical. While some experts believe that token prices will eventually fall, this has yet to happen.

The oversight needed for AI-generated tasks

On top of the costs of using AI tools, LLMs still need oversight from humans. One study from Harvard Business Review found that AI tools actually increased workloads for employees and managers. In the study, engineers reported spending more time reviewing, editing, and guiding AI-generated or AI-assisted work produced by colleagues.

Even if companies consolidate their workforces, they’ll still need humans to go through the work that an AI tool produces.

A recent analysis from the workforce analytics and productivity-tracking software company ActivTrak also found that workloads increased for employees. According to the study’s data, multitasking rose by 12% and weekend work increased by more than 40%.

The missing “human touch”

While Anthropic’s report suggested that customer service reps are at high risk of AI displacement, there’s evidence of a growing backlash against chatbots.

In a recent survey from the cloud hosting provider Kinsta, 93.4% of consumers said they prefer interacting with a human over AI. 71% of respondents also said they encountered situations where AI struggled with complex issues.

Some companies that tried replacing human customer service reps with LLMs have had to backtrack.

Between 2022 and 2024, the fintech company Klarna fired around 700 customer service assistants and replaced them with an AI assistant built by OpenAI, claiming that LLMs could handle their workers’ tasks. However, the plan backfired, with customers growing increasingly dissatisfied. The company eventually began rehiring human customer service assistants in 2025.

Anthropic’s report left many white-collar workers shaking in their boots. Fortunately for them, in most professions, it seems to still be too hard to eliminate workers and let AI tools reign free—humans are still very much needed.

But AI’s impact on the working work also shouldn’t be understated.

ActivTrak’s analysis found that 80% of employees now use AI tools at work, up 52% from 2 years ago. The average time workers spend using AI tools also increased eightfold. 4 Rather than flat-out replacing humans, it’s likely that using AI will become a necessary part of many white-collar jobs, especially those at the top of Anthropic’s list. This process is already well underway.

That also doesn’t mean that all workers are safe. Even though not all of the cost savings have materialized yet, firms are still increasingly moving toward consolidation. Dr. Matt Hansan, a Ph.D. economist and founder of the AI Humanist Movement, told Side Hustles that we’re seeing the “early phase of organizational redesign.”

“Firms are looking to create AI-augmented workflows, which will reduce the demand for more routine roles while increasing demand for people who can supervise, integrate, and manage AI systems,” he said. “One AI-enabled employee can often do the work that previously required several people.”