AI workplace paradox shows higher productivity and higher anxiety

At a glance:

• Anthropic’s survey of 81,000 Claude users finds one-fifth worry about displacement, with those in AI-exposed roles (such as developers and IT workers) voicing concern three times as often as peers in less-exposed positions.
• Highest-paid occupations report the largest productivity gains from AI, with 48% citing the ability to perform new tasks, 40% citing faster work, and just over 10% citing improved quality.
• Analysts warn automation of entry-level tasks (documentation, basic coding, routine analysis, structured support) risks narrowing early-career pathways and may leave enterprises short of mid-level experts years later.

Anxiety and adoption collide in AI-exposed roles

Workers are navigating a conundrum: even as AI dramatically accelerates their productivity, many fear it will ultimately displace them. According to a new survey from Anthropic, employees in roles most likely to be automated — including computer programmers, data entry keyers, market researchers, software quality assurance analysts and testers, information security analysts, and computer user support specialists — recognize their precarious position. Yet they readily adopt tools that could take their jobs and see first-hand how effectively those tools perform.

This measurement diverges from conventional displacement studies. Thomas Randall, research director at Info-Tech Research Group, noted that macro reports from Goldman Sachs, the International Monetary Fund (IMF), and the World Economic Forum (WEF) typically estimate what share of existing job tasks AI could theoretically perform in the future. By contrast, Anthropic is measuring qualitative experiences of workers in the present, which reveals how people are navigating this landscape in real time.

Productivity gains skew toward high-value work

Anthropic’s survey gauged the “visions and fears” of 81,000 Claude users and cross-referenced these views with platform usage data that flags jobs as more exposed when associated tasks are significantly performed in work-related contexts on Claude and occupy a larger share of a role. One-fifth of respondents expressed concern about displacement, noting their job or aspects of it are being taken over by automation. Those in the most exposed jobs voiced worry three times as often as peers in less-at-risk positions, and early-career respondents were also more nervous than others.

At the same time, the highest-paid occupations reported the largest productivity gains when using AI. These gains are most notable in the ability to perform new tasks, cited by 48% of users, while 40% said the technology helped speed up their work and a little more than 10% said it improved quality. Sanchit Vir Gogia, chief analyst at Greyhound Research, said enterprise usage is “actually quite consistent,” with teams applying AI where information is abundant and time is limited — such as drafting documents and code, summarizing content, responding to customer queries, and navigating internal systems.

When faster work invites harder work

Not everyone finds AI makes their jobs easier or faster. In some cases, people feel it makes their work harder; Anthropic noted that project managers are assigning tickets for issues that are much more difficult to solve. Gogia agreed that even when tasks become easier, scope and responsibilities expand and roles can absorb adjacent tasks, resulting in a redistribution of effort rather than a reduction of effort. Faster generation raises expectations on quality, and more output feeds into decision pipelines that are already constrained; in some cases, the system becomes heavier, not lighter.

Entry-level pathways erode despite delayed enterprise impact

The market is rewarding those who can integrate AI into complex workflows to do more, faster, and often with better outcomes, yet the most exposed tasks — documentation, basic coding, routine analysis, and structured support work — often sit at the base of the experience ladder. Automating these tasks reduces the urgency for companies to hire early-career workers, and what begins to erode is not the job but the path into the job. This can have a delayed impact: enterprises may not realize until years later that they lack enough mid-level experts because they did not bring enough people in at lower levels.

As AI plays a greater role in the workplace, Gogia said, there must be a conscious effort to rethink how people enter and grow. New pathways need to be created, and they need to be deliberate. Sentiment moves faster than structural change: workers feel the shift almost immediately, but organizations take longer to adjust hiring, redesign roles, and rethink workforce structures. Acceleration occurs in unchanged systems that still carry the same dependencies, approval chains, and coordination challenges, which is why expectations can become misaligned.

Redesigning work and measurement for an AI era

Leaders must approach the shift with intentional design, Gogia emphasized, clarifying how work is expected to change, what will be enhanced, what will reduce, and where development should focus. Baselines are moving: roles may begin to look oversized as what used to be considered a full day’s work begins to look like half a day’s work, and what used to be considered efficient begins to look average. AI is changing how work is done, but more importantly, it is changing what work expects from people.

Info-Tech’s Randall pointed out that workers who experience AI expanding what they can do by performing tasks previously outside their competence relate to AI more positively than those who experience it as doing their existing job faster. Tech leaders should therefore design AI deployment around capability extensions. Along with goal setting, managers need support to set expectations and interpret strategy; without it, even the best tools will fall short. Measurement must evolve to track quality, sustainability, and capability development over time. What we are witnessing is not a sudden disruption, Gogia said, but a gradual shift that is becoming impossible to ignore.