Will I Need to Reskill Because of AI?

Will I need to reskill because of AI? A 2026 SHRM survey of 5,875 US workers put that exact question to people already doing a job, not to a forecasting model, and the honest answer is: it depends far more on how visible AI already is in your field than on how exposed your specific role turns out to be.

What the survey actually asked

SHRM's "Navigating AI in the Workplace: 2026" report surveyed 5,875 US-based workers ages 18 to 90 in March and April 2026 through a third-party online panel, screening out anyone self-employed, retired, or working as an independent contractor so the sample reflects people currently employed by someone else. Several questions were shown to a random subset of 2,030 of those respondents rather than the full sample, a detail the report states directly instead of folding every figure into one topline number.

So will I need to reskill because of AI? It depends on the field

Across the full sample, 45% of workers said they are likely to need to reskill due to AI's potential to perform aspects of their current tasks and responsibilities. That average hides a wide spread by academic background: 66% of engineering and technology graduates expect to reskill, against 51% from business, public administration or communication fields, 51% from education and interdisciplinary studies, 37% from social sciences, 36% from health and medicine, and 34% from arts and humanities. SHRM's own reading is that workers in technical fields expect more reskilling not necessarily because their jobs are more exposed than anyone else's, but because AI's presence in those fields is already easier to see day to day.

The security paradox

The same survey holds two numbers that look like they should not sit together. 69% of workers said they feel secure in their role, even though 46% believe AI can already perform a large portion of their current tasks. Feeling secure and believing a large share of your own task list is already automatable answer two different questions: one is about a specific employer's near-term plans, the other is about what a tool is technically capable of doing. The gap widens by seniority too. Directors and above reported improved career prospects at twice the rate of individual contributors, which tracks who typically decides how AI gets used on a team rather than who performs the underlying task.

What to do regardless of the number for your field

A field-level percentage is not a verdict on one person's role, and self-rated skill and exposure track real ability only loosely, so a worker's own sense of how exposed they are is a weak instrument on its own. What holds up regardless of a reader's field or academic background is the same practical move this site keeps coming back to: build and show evidence of the skills a model cannot absorb now, rather than waiting for a survey question to settle it. Proof beats a listed skill in a hiring process, and that holds whether the immediate question is getting hired or staying useful inside a current role. The six-domain check is one way to get a specific, checkable read on where a reader already stands, instead of relying on a feeling of security or a general sense of exposure.

The honest limit

This is one survey of US workers taken over a two-month window, not a worldwide measurement of who genuinely needs to reskill and who does not. Self-reported reskilling expectations are not the same as verified skill gaps: the finding describes what workers believe about their own tasks, not an independent audit of any specific job. Read the 45% as a signal of how visible AI already is inside a field, not as a prediction for a single reader's role.