There is a narrative about AI and work that has been gaining traction: that generative AI is not eliminating jobs so much as reshaping them.
A recent field study published in the Harvard Business Review (2026) finds that organizations are redesigning roles around generative AI — augmenting existing tasks, changing the structure of work, and upskilling knowledge workers rather than laying them off. The study describes a pattern of role transformation, not headcount reduction.
It is a more optimistic story than the one most people have been hearing. And it is only half the picture.
The other half is coming from operational roles — customer service, data entry, administrative support. A large-scale SHRM survey of the U.S. workforce (2026) identifies these positions as the highest-displacement-risk clusters. A preprint study (2025) projects that customer service representatives face an 80% automation risk by 2025, and 7.5 million data entry clerk roles are projected to be eliminated by 2027.
These two shifts are happening in parallel. Neither one feeds into the other. One displaces workers from the bottom of the organizational hierarchy. The other reshapes them in place. And the total effect on headcount is not yet clear.
What the evidence actually says
The SHRM survey is methodologically interesting because it does not measure actual terminations. It asks workers to rate their own vulnerability to automation and AI-driven job displacement. The result is a perceived risk map — useful, but not a proxy for hiring freezes or layoffs. The report itself, sponsored by SHRM, makes this distinction.
The HBR field study is stronger evidence. It is based on actual organizational data — task-level observations, not self-reports. The finding is that generative AI is changing the content of white-collar work: what tasks people do, how they do them, which skills become more valuable. But it does not directly measure whether this reshaping preserves the total number of positions.
The projection study from 2025 is the oldest and least reliable of the three. Its figures (80% automation risk, 7.5 million roles) are forecasts, not realized outcomes. They conflate technical feasibility with organizational adoption — a mistake HR leaders make when they treat a 2025 model as a current operational reality.
The tension between them
Here is what the three sources together suggest, without claiming they prove it:
Routine operational roles are facing immediate, high-volume automation pressure. The work is standardized, measurable, and increasingly handled by AI systems. Displacement here is structural — these roles can be eliminated more easily than knowledge work roles.
White-collar knowledge work is facing gradual, task-level redesign. AI is not eliminating these roles yet, but it is changing what they require. Skills that matter today may not matter in three years. The work is being reshaped, not removed.
The gap between them is the problem. Displaced operational workers are unlikely to move into white-collar roles that have been reshaped rather than reduced. The skills required are different. The organizational status is different. The pay is different. And there is no evidence that reshaping knowledge work creates room for operational workers who are being displaced.
This is the tension the HBR study does not address: if routine roles are vanishing while white-collar roles are being reshaped in place, what is the net effect on the size and structure of the workforce?
What management can actually do
The research does not give us a simple playbook. But it does suggest several things that go in the right direction:
Audit operational workflows for automation exposure now. The projection study is old, but the direction is established. Customer service, data entry, scheduling, and reporting are the highest-risk clusters. Organizations that wait for the displacement to happen before preparing for it will be managing a crisis, not a transition.
Focus on role crafting, not headcount defense. The HBR study finds that role redesign is the dominant outcome. That means the question for managers is not “how many people do we need” but “what will these people do.” Up-skilling operational staff for tasks that AI cannot easily automate is more valuable than trying to preserve roles that will inevitably change.
Distinguish perceived risk from realized risk. The SHRM survey is useful for understanding how workers feel about their job security. It is not a reliable predictor of actual layoffs. Treating perceived vulnerability as a hiring freeze signal would be a mistake. But treating it as irrelevant would be another.
Watch for the acceleration point. The HBR study describes “early light” evidence. Reshaping work today does not guarantee job security tomorrow. If organizational adoption of AI accelerates, the difference between “reshaping” and “eliminating” may become much smaller.
What this doesn’t tell you
These studies are early. The SHRM survey measures perception. The projection study measures theoretical feasibility. Only the HBR field study measures what actually happens in organizations — and it describes a pattern that may shift as diffusion deepens.
None of them measure the interaction between the two shifts. We do not know whether AI-driven task redesign in white-collar work will eventually reach the same scale as operational automation. We do not know whether displaced operational workers will find new roles in the reshaped knowledge economy. We do not know if the current pattern — displacement at the bottom, reshaping at the top — is a stable equilibrium or a transition phase.
What we do know is that both shifts are real, both are happening in parallel, and neither cancels the other out. Management practices that focus on only one of them are working with an incomplete picture.
The question is not whether AI is reshaping work or displacing workers. The question is what happens when both are happening at once.
Sources:
- “Automation, Generative AI, and Job Displacement Risk in U.S. Employment,” SHRM (published January 2026)
- Research: How AI Is Changing the Labor Market, Harvard Business Review (published March 2026)
- “AI Job Displacement Analysis (2025-2030),” SSRN preprint (published June 2025)