AI Reskilling Jobs: 92% of HR Leaders Say Staff Can Retrain — Yet Firms Still Hire Outside

AI reskilling is the answer most employers believe in, and the one most of them are not using. New research published on 4 November 2026 found that 92 per cent of human resources professionals believe their own employees could be retrained for AI-related roles, yet more than half of their organisations still hire from outside when those roles open up.

The findings come from Careerminds, a workforce solutions provider, in a white paper called The AI Redeployment Gap. The research, carried out with Censuswide between 15 and 26 May 2026, surveyed 1,142 human resources professionals and people managers in the United States, at organisations with 500 or more employees. It describes a gap between what employers say about their people and what they actually do when AI creates new work.

AI reskilling: belief is nearly universal, action is not

The headline numbers are striking. Ninety-two per cent of HR professionals and 88 per cent of people managers said employees inside their organisations could be reskilled for the roles AI is creating. Yet 51 per cent of HR professionals and 47 per cent of people managers said their organisations hire externally for new AI-related positions, compared with just 42 per cent in each group who redeploy existing staff.

The gap is widest where AI hiring is hottest. In technology and telecom, external hiring outpaces redeployment by 15 percentage points among both groups. That is the sector where AI roles are multiplying fastest, and where existing staff are least likely to be moved into them.

Part of the problem is that nobody owns the decision. Sixty-three per cent of HR professionals said HR is responsible for AI workforce planning, but only 24 per cent of managers agreed. When two groups disagree that sharply about who is in charge, the default is to post a vacancy and hire from outside. The 2,965 AI agent jobs open in November 2026 show how strong that outside demand is.

Why companies cannot see who to retrain

A second problem is visibility. Forty-three per cent of HR professionals and 34 per cent of people managers said their organisations had limited visibility about who should even be reskilled. Companies hold records of job titles and reporting lines, but they rarely hold a clear map of the skills their people actually have, so managers cannot see that the person in the next team already has most of what an AI role needs.

That blind spot is expensive. The white paper draws on analysis from the Josh Bersin Company, based on roughly half a million hiring records, which found that hiring externally takes longer and shows lower retention than moving someone internally. An external hire needs months to learn how the organisation works, while an existing employee already knows the customers, the systems and the unwritten rules.

The pattern also feeds the churn the jobs market is already seeing. Companies that cut roles because of AI are now bringing some back: a Robert Half survey found 32 per cent of employers that eliminated a role because of AI later rehired for the same or a similar role, as the AI layoffs reversal figures show. Hiring outside for AI work while laying off the people who could have done it is the same mistake made twice.

What the AI reskilling gap means for workers

For workers, the gap is good news and bad news at once. The good news is that employers are hiring for AI roles from outside, so the door is open to anyone who can show the skills. The AI training jobs open to software engineers are one example of the new work, and the pay for that kind of role is often published openly on the posting itself.

The bad news is that staying put is no longer a safe strategy. If your employer believes you could be reskilled but still hires outside, waiting to be noticed will not work. The practical move is to name your skills in terms the hiring manager recognises: which AI tools you have used, what you built with them, and what went wrong and how you fixed it. That evidence matters more than a course certificate, because it shows judgement, which is the part of the work employers say AI cannot carry.

For people already in work, the survey points to a direct question worth asking a manager: what AI skills is the team short of, and what would it take to be considered for the next AI role that opens? If the answer is vague, the visibility problem is real in your organisation, and the safer path is to build the skills in the open, where a hiring manager outside the company can see them too.

Anyone approached about an AI role should apply through the employer's own careers page and never pay for a job, a training place or a placement. The United States Federal Trade Commission warns jobseekers never to pay for a job and to ignore unexpected job messages on WhatsApp or Telegram.

At a glance	Details
Research	Careerminds white paper, The AI Redeployment Gap, published 4 November 2026
Survey	Censuswide, 15 to 26 May 2026; 1,142 US HR professionals and people managers at organisations with 500 or more employees
Believe staff can be reskilled for AI roles	92% of HR professionals; 88% of people managers
Hire externally for new AI roles	51% of HR professionals; 47% of people managers, against 42% in each group who redeploy existing staff
Technology and telecom	External hiring outpaces redeployment by 15 percentage points
Who owns AI workforce planning	63% of HR professionals say HR; only 24% of managers agree
Visibility gap	43% of HR professionals and 34% of managers report limited visibility about who should be reskilled
Cost of hiring outside	Takes longer and shows lower retention, based on analysis of roughly half a million hiring records

Sources: PR Newswire at https://www.prnewswire.com/news-releases/new-research-92-of-hr-professionals-believe-employees-can-be-reskilled-for-ai-jobs-yet-external-hiring-outpaces-redeployment-302927992.html, publishing the Careerminds research on 4 November 2026; and Careerminds at https://careerminds.com/resources/redeployment-gap on the white paper's key findings.