Key Takeaways
The most dangerous question when selecting reskilling candidates is, “Which jobs will disappear?” This question quickly creates fear, but it is too crude as an execution standard. A more useful question is: “Which work is shrinking, which work is growing, and which roles can people move into?”
SHRM’s 2026 AI in HR report, based on a survey of 1,908 HR professionals, states that respondents in organizations where AI has been deployed reported frequent upskilling or reskilling opportunities at 57%, changes in job responsibilities at 39%, new roles at 24%, and some job displacement at 7%. It also explains that the organizational impact of workplace AI is 5.7 times more likely to shift job responsibilities and three times more likely to create new roles than to displace jobs.
Therefore, reskilling candidates should be found not among “people in jobs that will disappear,” but among “people whose mix of work is changing, who can move into adjacent roles, and whose learning and placement can be designed together.”
Reskilling Candidates Come From “Changing Work,” Not “Disappearing Jobs”
Looking only at job titles can cause organizations to define reskilling candidates too broadly or too narrowly. For example, within the same customer support role, simple inquiry responses may be automated, while complex issue coordination, customer complaint analysis, and AI response quality review may become more important. If the entire job is treated as a risk group, the organization misses the work that remains and the roles that grow.
SHRM’s survey scope is a global HR professional sample, and its industry and job-group composition may differ from Korean companies. Even so, the figures—39% changes in job responsibilities, 24% new roles, and 7% job displacement—show that reskilling should start from work change, not job elimination. Candidate selection must begin by mapping changes at the unit-of-work level rather than by job title.
The First Criterion Is Not Automation Potential, but the Judgment Work That Remains
Identifying what AI can do is not enough to select reskilling candidates. Even when work has high automation potential, judgment work that humans need to perform better may remain around it. This is why SHRM’s point that workplace AI is 5.7 times more likely to shift job responsibilities than to displace jobs matters.
The first criterion, then, is not “what will be automated,” but “what judgment remains after automation.” If repetitive data entry decreases, exception handling and quality checks may remain. If report drafting becomes faster, data interpretation and decision support may become more important. If counseling responses are automated, customer issue analysis and service-improvement proposals may grow.
The Second Criterion Is Adjacent Roles Into Which People Can Move
Reskilling does not end with training. There must be adjacent roles into which people can be placed. TalentLMS’s 2026 L&D Report states, “Training builds skills, but mobility builds futures,” emphasizing the need to connect internal career paths with learning. At the same time, it reports that 44% of HR managers prioritize external candidates over internal employees for new roles.
This figure shows how important internal mobility paths are in selecting reskilling candidates. If a company says it will develop internal candidates but fills new roles through external hiring, reskilling loses motivation. When selecting candidates, HR should examine the distance between current skills and the target role. If the distance is too small, the case is closer to upskilling; if it is too large, short-term reskilling may not be realistic. It is practical to begin with groups where adjacent roles are visible.
The Third Criterion Is Learning Potential and Placement Potential Together
Reskilling candidate selection can fail if it looks only at willingness to learn. Learning without placement potential becomes an ambiguous promise for both the individual and the organization. TalentLMS reports that 64% of HR managers choose upskilling or reskilling current employees to address skills gaps, 62% use AI to automate work, and 57% rely on hiring external specialists.
TalentLMS also states that half of companies are restructuring roles or responsibilities, and 29% are eliminating positions that rely on outdated skills. These findings should be interpreted differently depending on each company’s situation, but one thing is clear: reskilling is not a training program. It is a workforce-strategy choice among automation, role restructuring, external hiring, and internal mobility.
Practical Matrix: Four Questions for Separating Reskilling Candidate Groups
First, are shrinking work and growing work within the current job clearly distinguished? Second, is there adjacency between current skills and the target role? Third, can learning produce a verifiable output within three to six months? Fourth, is there a project or role available after training?
Using these four questions makes execution priorities clearer. A group with major work change, adjacent roles, and high placement potential is a first-priority reskilling target. A group with major work change but low role adjacency or placement potential needs longer-term transition or separate workforce planning. A group with limited work change whose main need is better current-role performance is closer to an upskilling target than a reskilling one.
Since CompTIA reports that 83% of organizations place a high priority on addressing skills concerns, many companies face similar questions. What matters is not labeling all employees as reskilling targets at once, but distinguishing survey population, job groups, work change, and placement potential—and starting with a small pilot.
What HR Should Watch Next
Selecting reskilling candidates is not a process of classifying individuals as risk groups. It is a workforce-strategy decision that looks at work change and role-transition potential through data and checks whether the organization can actually open mobility paths. HR must look at work, not job titles, and consider placement potential alongside willingness to learn.
The next article will cover the skills data design that makes this judgment possible. Before adopting an LMS or AI learning platform, organizations need to define what skills data should be accumulated and validated, and how.
