What workers expect from AI, and what that asks of their managers

In short: most US workers are more worried than hopeful about AI at work, employers expect almost 40 percent of job skills to change by 2030, and the best data so far shows no economy-wide job loss but real pressure on early-career workers in the most exposed roles. Closing the gap between those facts is a management job: a plan for people, training for judgment, and protection for the entry-level pipeline.^1,2,3,4

Two surveys from 2025 describe the same moment from opposite ends. In February, Pew Research Center reported that 52 percent of US workers were worried about the future impact of AI in the workplace, 36 percent were hopeful and 33 percent felt overwhelmed. Asked about their own prospects, 32 percent expected fewer job opportunities in the long run and only 6 percent expected more. The survey, of 5,273 employed adults in October 2024, also found how little AI was yet part of daily work: 16 percent said at least some of their work was done with AI, and 63 percent said they did not use it much or at all.¹

In January, the World Economic Forum’s Future of Jobs Report 2025, based on a survey of more than 1,000 employers, projected 170 million jobs created and 92 million displaced by 2030, a net gain of 78 million, and estimated that nearly 40 percent of the skills required on the job would change over the period.²

Employers see net growth and a skills transition. Employees see uncertainty about their own role. Both can be true, and the distance between them is where adoption programs succeed or stall.

What the job data shows so far

Surveys measure expectations. Two studies published in late summer and autumn 2025 measured what was actually happening, and they point in usefully different directions.

The Stanford authors found the adjustment showing up in headcount rather than pay, and concentrated in occupations where AI tends to automate tasks rather than augment the people doing them.⁴ The Yale team, for its part, says better data is needed and that it will update its analysis regularly.³ Read together, the studies say: the economy is not losing jobs to AI at scale, but the first rung of the ladder in exposed fields is getting narrower.

What leaders are planning

Microsoft’s 2025 Work Trend Index found 82 percent of leaders calling 2025 a pivotal year to rethink strategy and operations, and leaders expecting their teams to be training AI agents (41 percent) and managing them (36 percent) within five years.⁵ In the WEF survey, 85 percent of employers planned to prioritize upskilling, half planned to reorient their business in response to AI, and about four in ten expected to reduce their workforce where AI can automate tasks. The WEF estimates that 59 of every 100 workers will need training by 2030.^2,6 That is a substantial change in what a manager’s job contains, announced to a workforce that is more worried than hopeful.

Four things that close the gap

1. Say what the plan is for people. Worry grows in silence. Employees want to know which tasks change, whether roles change, and how they will be trained. A short, specific answer beats a long reassuring one.
2. Train for judgment, not only for prompts. The skill that lasts is knowing which tasks the tools do well in your field, which they do badly, and how to check the difference. The Deloitte refund is what happens when that check is missing.
3. Protect the entry-level pipeline. If AI absorbs the routine tasks junior staff used to learn on, decide deliberately how they will now learn: rotations, shadowing, review of AI output as a training task. Today’s juniors are the only source of tomorrow’s experienced staff.
4. Make it safe to say “this did not work.” Early users find the failures first. If reporting a bad output feels like admitting incompetence, you will hear about failures from customers instead.

For your own career

The same evidence reads differently from the individual’s side. If nearly 40 percent of the skills in a job are expected to change by 2030, the safest position is to be the person who understands both the work and the tools: which steps can be delegated, which must be checked, how to supervise an agent that acts on its own (the Replit incident shows why), and how to measure whether the change helped.² That is a management skill, and it is learnable; it is what the AIMP certification is built to teach and to show an employer.

A practical start: pick one recurring task in your week, try the approved tools on it for a month, and write down where they saved time and where they created rework. In a year of surveys about what AI might do to jobs, a month of your own evidence is worth more than any forecast.