BLOG

The Dual Trap

Posted · Add Comment

How AI Layoffs and Token Shock Are Defining the CFO’s AI Workforce Strategy in 2026

At an enterprise event in early June 2026, OpenAI CEO Sam Altman acknowledged something that caught corporate America mid-sip: cost management has become the second most common complaint he hears from enterprise customers. A meme was circulating in boardrooms. It went something like this: The company spent its entire 2026 budget in Q1. Can you make it more efficient?

That is not a punchline. That is a boardroom in distress.

The distress compounds when CFOs look at the other side of the ledger and realize they are also paying to undo decisions made eighteen months ago. The companies that rushed to replace humans with AI are now contending with two simultaneous crises. Most CFOs have started to feel one. Fewer have fully reckoned with both at once. That gap is now the defining challenge in CFO AI workforce strategy.

Trap One: The AI Boomerang

Start with Klarna. In early 2024, the Swedish fintech company made headlines for announcing its AI assistant could do the work of 700 customer service agents. The press release was triumphant. By mid-2025, CEO Sebastian Siemiatkowski told Bloomberg the quiet part out loud: “Cost was too predominant as an evaluation factor. The result was lower quality, and that is not sustainable.” Rehiring began. The story that was supposed to demonstrate AI’s triumph over human labor became one of the clearest illustrations of why full replacement strategies fail.

Klarna is not an outlier. It is the case study. And the data behind it is now substantial.

  • Nearly one-third of companies that cut staff due to AI have already rehired for those exact roles (Robert Half, survey of 2,000 hiring managers, 2026).
  • Two-thirds of companies that conducted AI-driven layoffs are already rehiring some of those workers. Among them, 35.6% brought back more than half the eliminated roles, and 52% did so within six months (Careerminds, survey of 600 HR professionals, February 2026).
  • 55% of executive decision-makers who replaced employees with AI now regret the decision (Forrester, Predictions 2026: The Future of Work).
  • By 2027, 50% of companies that attributed headcount reductions to AI will rehire staff to perform similar functions, often under different job titles (Gartner, February 2026).

The financial case for AI-driven layoffs has also proven far weaker than advertised. Only about one-quarter of companies that executed AI-driven cuts ended up financially ahead. Nearly one-third spent more on restaffing than they saved. Another 42% broke even. Those figures do not account for the harder-to-quantify costs: institutional knowledge that walked out the door, team morale that quietly collapsed, and customer trust that took months to rebuild.

The headline number from that same Careerminds research may be the most clarifying statistic in the entire dataset: only 8.4% of HR leaders said their AI restructuring delivered what was promised and would repeat the process unchanged. 91.6% would do it differently.

Executive watching a returning boomerang composed of employees, knowledge, and business relationships, symbolizing the AI layoff boomerang effect and the hidden value of human capital.

Treating AI as a workforce solution rather than a workforce tool is the strategic miscalculation every boomerang rehire makes visible. It is also the one 91.6% of HR leaders say they would not repeat.

Trap Two: Token Shock

While the boomerang has received most of the attention, a second crisis is running parallel and has received far less. It is unfolding in the same boardrooms, on the same CFO’s desk, often in the same quarter.

Uber’s CTO Praveen Neppalli Naga confirmed to The Information that the company burned through its entire 2026 AI budget in four months. The driver was the adoption of Claude Code across 5,000 engineers, at individual costs running $500 to $2,000 per engineer per month. No budget model had anticipated the pace of adoption. Uber’s COO later said publicly that without a clear line between AI spend and product improvements shipped, “that trade becomes harder to justify.”

Uber is not alone.

  • Microsoft quietly canceled most internal Claude Code licenses and redirected engineers to a less expensive alternative.
  • Amazon scrapped its internal token leaderboard and directed staff to stop using AI “for its own sake.”
  • Walmart ended its policy of giving workers unlimited token access.

The structural dynamic underneath these stories is the one CFOs have not fully absorbed: token costs have no natural ceiling. Payroll did. A company that replaced a salary line with a token line traded a predictable operating cost for an unpredictable one, denominated in units most finance teams still cannot audit at the function level.

Sam Altman’s own data frames the magnitude. AI token consumption has grown one million-fold in six and a half years. What was once the usage level of the world’s most extreme power user is now roughly the global per-person average. Altman expects it to grow another million-fold from here. For CFOs who set AI budgets based on current consumption rates, that trajectory is the risk.

Forrester Research adds a structural warning: as AI’s hype period ends, fewer than one-third of decision-makers can tie the value of their AI investments to their organization’s financial growth. 48% of CFOs now say they are ultimately responsible for ensuring AI delivers measurable value. Most of them do not yet have the tools or governance frameworks to do that.

Where the Two Traps Meet: The Cost of a Flawed AI Workforce Strategy

The boomerang and token shock look like different problems. They share a root cause.

Research consistently shows that AI handles roughly 60% of routine, structured work reliably. The remaining 40% requires nuance, contextual judgment, institutional knowledge, and the kind of human-to-human trust that does not reduce to a prompt. Companies that hit the boomerang trap did not model the 40%. Companies hitting token shock did not model unconstrained usage of the 60%.

The Robert Half survey of 2,000 hiring managers puts specific texture on the gap: 40% said AI could not replace institutional knowledge, 38% said they underestimated the need for human quality control, and 35% saw disappointing productivity gains. Those are not technology failures. They are leadership and judgment failures.

The Conference Board’s 2026 C-Suite Outlook Survey found that 38% of U.S. CEOs now identify AI as the leading factor that could negatively affect their businesses. Not positively. That represents a significant shift in executive sentiment from just eighteen months ago, and it reflects the reality that most AI strategies were built on assumptions that did not hold.

Both traps flow from the same strategic miscalculation: companies made workforce decisions based on what AI promised rather than what it had actually proven at scale. And they made those decisions without leaders in place who understood the difference.

The Companies That Got It Right Read the Situation Differently

Not every organization made this mistake. Some read the situation clearly from the start.

When IKEA’s AI chatbot Billie began handling 47% of routine customer service inquiries starting in 2021, the company faced a decision. It did not lay off 8,500 call center workers. It retrained them as remote interior design consultants, a role built around exactly the taste, relationship skill, and contextual expertise that AI cannot replicate. The remote consultation channel generated €1.3 billion in revenue in fiscal year 2022 alone and has continued to grow.

IKEA did not ask how many people it could replace. It asked what work humans should be doing that they currently were not. That is a fundamentally different leadership question, and it produced a fundamentally different outcome. Billie did not eliminate 8,500 jobs. It revealed them.

The difference between IKEA and Klarna is not the AI. It is the quality of the strategic judgment at the leadership level.

What This Means for the C-Suite

The question at the center of every CFO AI workforce strategy right now, whether quietly in audit committee or explicitly in board strategy sessions, is not technical: What is the right human-to-AI ratio in each function, and who in this organization decided that rigorously?

That is not a question a Chief AI Officer built around workflow automation can answer well. That archetype has already proven costly in too many organizations. The executive who can answer it is what Hager Executive Search calls the Platinum Knowledge Worker: the operational leader who understands what AI can and cannot do, who can run a function, motivate a team, and design workflows around genuine human-AI collaboration rather than replacement.

Finding that executive requires something most search processes are not designed to do: distinguishing genuine AI fluency from performed AI fluency. A candidate who has memorized the right answers about transformation and automation is not the same as a leader who has actually made the judgment calls that prevent a Klarna outcome or a token budget crisis.

That distinction is precisely what Hager’s Cognitive Search Methodology was built to make. Before the search launches, the organizational diagnosis has to be right. The role has to be defined around what the company actually needs, not what the press release describes. And the evaluation has to probe for the judgment that the market is now pricing at a premium.

For a deeper look at how to evaluate AI fluency in C-suite candidates before a search begins, see How to Hire an Executive Who Actually Understands AI.

The CFO’s Mandate for the Rest of 2026

A sound AI workforce strategy for CFOs starts here: A sustainable CFO AI workforce strategy treats AI spend as a controlled operating line, not a headcount replacement bet. The CFOs navigating the second half of 2026 successfully are the ones treating token budgets with the same discipline they once applied to headcount: by function, by model tier, by measurable output per dollar spent.

Before any further restructuring decision, the board should be able to answer five questions:

What AI system is in production and proven at scale in this organization?

Pilot programs and proof-of-concept demos do not count. The board needs evidence the system performs reliably under real operating conditions before workforce decisions are tied to it.

What are its error rates, and how does it handle the edge cases where human judgment is required?

Every AI system has a failure mode. The organizations that avoid the boomerang trap know exactly where theirs is and have a human positioned to catch it.

What is the rollback plan if AI performance degrades?

If the only plan is rehiring the people who were laid off, that is not a rollback plan. It is the boomerang already in motion.

How is institutional knowledge being protected, not just during the transition but in the years after?

Institutional knowledge that leaves with a laid-off employee rarely returns even when the employee does. Protecting it has to happen before the layoff, not after.

Is the organization measuring cost per output, or simply total spend?

Total spend tells you how much you used. Cost per output tells you whether it was worth it. Most CFOs currently only have the first number.

And there is a harder question underneath all five: does the organization have the leadership in place to hold both sides of this equation? The operational side and the AI governance side, simultaneously, without conflating them?

If the answer is no, that is an executive search conversation. And it is the conversation the CFO needs to have before the next budget cycle, not after it.

About Hager Executive Search

Hager Executive Search is a San Francisco-based retained executive search firm serving growth-stage and mission-driven organizations. Three-time Forbes-recognized among America’s Best Executive Search Firms, Hager combines management consulting rigor with retained search to diagnose organizational gaps before launching a search. In 2026, that means evaluating genuine AI fluency in C-suite candidates, not just functional pedigree. Learn more at hagerexecutivesearch.com.

Sources cited in this article:

 
 

READY TO BUILD A WINNING LEADERSHIP TEAM?

CONTACT US