A survey of 1,006 global executives, reported by Harvard Business Review, found that AI is already being cited as a reason for layoffs. That is not surprising in itself. What is notable is that most of those reductions appear to be anticipatory — driven by the belief that AI will soon replace work — rather than by measured proof that AI has already replaced it.
The difference between potential and performance
There is a meaningful distinction between “we have automated X and no longer need these roles” and “we believe AI will soon automate X, so we are reducing headcount now.” The first is a restructuring with evidence. The second is a bet on technology that may or may not pay off, made at the expense of people who were doing real work.
The risk is obvious. If the promised productivity gains do not materialise on schedule, the firm is left with fewer staff, unhappy survivors and an AI transformation that still has to be delivered. Anticipatory layoffs can damage the very operational capacity needed to make AI succeed.
Why firms jump the gun
Pressure from boards, investors and competitors creates a strong incentive to act fast. AI announcements move share prices and shape external perception. In that environment, headcount reductions can become a signalling device: a way to show seriousness about AI, even when the technology is not yet doing the work.
There is also a planning fallacy at play. It is easier to imagine a future in which AI handles a task than to do the integration, change management and quality assurance required to make it real. The gap between imagination and implementation is where bad workforce decisions get made.
A better sequence
The more durable approach is to build the human-AI operating model before you change the workforce. That means:
Map the work before you map the headcount. Break roles into tasks. Identify which tasks are genuinely automatable today, which are augmentable, and which still need human judgment. Headcount decisions should follow that map, not precede it.
Run parallel operations. Keep the existing team in place while the AI system proves itself at scale. Measure error rates, throughput and customer outcomes against the baseline. Only once the new model is demonstrably better and stable should you consider structural staffing changes.
Invest in redeployment, not only redundancy. In many firms, AI creates new roles — prompt engineering, quality assurance, workflow design, AI governance — faster than it removes old ones. Retraining existing staff is usually cheaper and less risky than hiring from scratch.
Governance and trust
Anticipatory layoffs also carry governance risk. Regulators and unions are increasingly scrutinising AI-related workforce decisions. Firms that cannot show a clear operational basis for reductions may face legal challenge, reputational damage and internal resistance that slows the wider programme.
Trust is a practical asset in AI adoption. Employees who believe AI is being used to make their work better will engage with it. Employees who believe it is being used to remove them will resist, hide errors and undermine the rollout.
The bottom line
AI will reshape workforces. But reshaping them before the technology has proven its case is a strategy built on hype. Build the operating model, measure the gains, then adjust the workforce. The order matters.