Jesse Jacoby
In 2021, Italy committed €4.4 billion to a labor market program built in large part around reskilling. The courses offered through public employment agencies were short and free, and many were aimed at occupations with real shortages, especially in IT. Two years later, only about one in ten jobseekers had started any training at all.
A team of economists that included Harvard Business School’s Raffaella Sadun set out to learn why. They surveyed roughly 1,100 unemployed workers in Italy and asked them to choose among training options: reskilling into an IT assistant role, reskilling into a construction technician role, or a generic Microsoft Office course for the work they already did. Both reskilling targets were desk jobs in high demand, and both paid better than most respondents assumed. Participants were willing to pay out of their own pockets, about €3.50 per training hour, to reskill into IT. They would have to be paid about €5.30 an hour to train for construction, and those who did not see themselves in that line of work needed close to €10.1
Wages, job prospects, and the effort of learning all mattered at the margins. The factor that explained the most was whether the job matched the person’s sense of who they were. At the Milan employment center that recruited part of the sample, a handful of the 512 jobseekers went on to take IT courses afterward. None took courses to move into construction.
That finding should concern every organization with an AI workforce plan, because almost all of those plans rest on the opposite assumption.
The Catalog Assumption
The scale of the coming transition is no longer speculative. The McKinsey Global Institute estimated in September that about 11 million U.S. workers in declining occupations may need to move into different occupations by 2035. That would require roughly 770,000 people a year to switch occupational groups, about three and a half times the historical pace. Only about one in seven of those workers has what McKinsey calls a direct pathway to new work. The rest face routes that are winding or not yet paved at all.2
Most enterprises are responding with what they already own: a course catalog and a dashboard that counts enrollments. The design logic underneath is economic. If the training is free, the destination pays more, and the path is clearly marked, rational people will take it. When they do not, the diagnosis is awareness, and the fix is another campaign.
The Italian evidence says the diagnosis is wrong. The courses were free. The destinations paid more. The barrier sat somewhere a catalog cannot reach. As Sadun put it, “People don’t see themselves as a bundle of skills; they see themselves as having a role in society.”
The Adjacency Nobody Maps
Skills-based workforce planning maps the adjacency between skill sets. A claims processor shares most of the skills of a data quality analyst, so the system flags a pathway. The claims processor is evaluating a different adjacency entirely: whether the person who does that job is someone they recognize, respect, and want to be.
The study put numbers on how much this matters. On a ten-point scale, perceived identity fit averaged 5 for the IT role and 3.7 for construction, and respondents associated even the desk-based construction role with lower social status, a perception that better pay did not dissolve. The gap in subsidies tells the same story from another angle: people with above-average fit for construction needed about €3.80 an hour to take the training, while people with weak fit needed close to €10. The course and the job were identical. The only thing that changed was whether the person could see themselves in it.
Inside a company, the identity cost is usually steeper than it looks on the org chart. A twenty-year supervisor asked to become a workflow designer has to stop being the person everyone calls when something breaks, and become a novice in front of the people they used to lead. Learning the software is the easy part. That is a loss of standing, and people price losses of standing very heavily. A catalog entry titled “Introduction to Process Automation” does nothing to offset it.
Plain Facts Beat Inspiration
The researchers tested two ways of shifting the decision. One group watched a short video with plain facts about pay and job openings in the target occupation. Another watched a growth mindset video about a woman who had successfully changed careers.
The fact-based video raised willingness to reskill by about 11 percent, and it was followed by a real rise in IT training enrollment at the Milan employment agency. The gap it closed was real: three quarters of respondents had undervalued wages in these occupations, and nearly 40 percent had underestimated demand. The inspirational video changed few minds.
Most corporate reskilling communication is the second video. A keynote on the future of work, or a success story from another function in another country. It is sincere, and it does not travel, because it gives people nothing specific to evaluate. What moves the decision is credible, local information, like what the destination role pays here and how the people who already made the move are doing. Most organizations have that data sitting in their HR systems and never publish it.
Describe the Destination Down to the Tuesday
People cannot step toward a role they cannot picture. Ask two managers what someone in a newly created role actually does on an ordinary Tuesday and you will often get two different answers, which means the role exists as a box and a list of competencies but not yet as work. A pathway that nobody can describe in terms of daily work is not a pathway, however neatly it appears in the skills platform.
The fix is concreteness. Write the role as a week of work, not a list of capabilities. Show the decisions it owns, the people it works with, and the problems it gets handed. Then put someone in front of the candidates who has actually done it, ideally a person who came from the same starting role. A peer who made the move is evidence. A poster child from another part of the business is a growth mindset video with a name badge.
Let People Try On the Role Before They Commit
Herminia Ibarra’s research on career change found that people rarely reinvent themselves by deciding first and acting second. They change by doing: taking on small projects, building new relationships, and revising their sense of self based on what they experience. Identity follows action far more reliably than it leads it.3
Most enterprise reskilling programs reverse that order. They ask people to commit to a new identity up front, enroll in a twelve-week program, and find out at the end whether the work suits them. For anyone with a mortgage and twenty years of credibility in their current role, that looks like a one-way door, and people treat one-way doors with appropriate caution.
Make the door two-way. Use rotations, shadowing, and part-time project assignments that let people do a slice of the destination work while keeping their current role. Make the return path explicit. The person who tries the new role for ninety days and decides it is not for them should come back with their reputation intact, or nobody else will try.
Protect Status Through the Move
The construction finding is a warning about status, and status is something leaders control more than they realize. Who goes first says more than any announcement. When the first cohort into a new role is made up of respected performers, the move reads as a promotion. When it is made up of people whose current jobs are being eliminated, it reads as a holding pen, and everyone watching draws the obvious conclusion.
Protect pay through the learning curve. Carry over titles or seniority where the role allows it. Give explicit credit for the institutional knowledge people bring, since it is often why a former supervisor designs better workflows than an outside hire. And measure the program by transitions rather than enrollments: how many people are doing the new work twelve months later, at what pay, and with what performance.
The Question Before the Budget
Reskilling budgets are going to be spent either way. The organizations that get a return on them will be the ones that treat the decision to move as a question of identity and status and design for it, rather than treating it as a question of access and answering it with another catalog.
Before the next workforce plan goes to the board, pick one role you expect to shrink and ask five of the people in it three questions. What would you be doing on a Tuesday in the role we want you to move into? What would it pay? Who here has made that move? If they cannot answer, another motivation campaign will not help. Nobody has shown them who they would become.
References
- Alexia Delfino, Andrea Garnero, Sergio Inferrera, Marco Leonardi, and Raffaella Sadun, “Unwilling to Reskill? Experimental Evidence from Real-World Jobseekers,” NBER Working Paper 34633, January 2026 (also HBS Working Paper 26-041). The study surveyed about 1,100 unemployed Italian jobseekers between April and September 2023, including a sample recruited through AFOL Metropolitana, the consortium of public employment agencies in metropolitan Milan. The paper identifies perceived identity fit as the most important individual-level factor in reskilling decisions and finds that a light-touch information intervention on occupational returns raised both stated interest in reskilling and real-world training participation. Background on Italy’s 2021 program, the 10 percent training start rate, the willingness-to-pay and subsidy figures, the 11 percent effect of the information video, the AFOL Milan enrollment outcomes, and the Sadun quotation are from Avery Forman, “Want Workers to Reskill? Show Them Who They Can Become,” Harvard Business School Working Knowledge, June 18, 2026.
- McKinsey Global Institute, “Workforce in Motion: Skills and Pathways to Future Jobs in the United States,” September 29, 2026. The report estimates that about 11 million workers in declining occupations (a range of 6 million to 16 million depending on the pace of automation) may need to change occupations by 2035, requiring about 770,000 occupational group switches per year, roughly 3.6 times the historical average. About 14 percent of transitioning workers have a direct pathway, 41 percent a winding pathway, and 45 percent an unpaved one.
- Herminia Ibarra, Working Identity: Unconventional Strategies for Reinventing Your Career (Boston: Harvard Business School Press, 2003). Based on in-depth study of professionals making career transitions, Ibarra argues that people change careers by experimenting with new activities, relationships, and narratives rather than by settling on a new identity in advance and then acting on it.
