Detroit, March 2026. John sits in a glass office on the twelfth floor, facing a whiteboard covered in red arrows. Across from him, the CFO waits for his decision. “Nine hundred positions. That’s the number the consultant gave us to match our competitors on operating costs. Generative AI handles the rest.”
John looks at the board. He thinks about his executive committee, about shareholder pressure, about the latest analyst report comparing his margin to three companies in the same sector, all three running massive reduction plans. He also thinks about a question nobody has raised in the last six months of meetings: what if AI wasn’t built to cut, but to build?
“Give me a week,” he finally says.
This scene repeats itself, in different forms, in hundreds of executive committees right now. The reflex that pushes John toward this dilemma is old, and it biases the decision before it is even made.
The knife bias
There is a simple neurological reason for this reflex. Cutting costs is measurable immediately, visible within a quarter, easy to present to a board. Building growth takes time, tolerates uncertainty, and requires accepting that the return will not be linear. The human brain, biased toward a certain, fast gain over a larger but delayed one, naturally pushes leaders toward the option that looks most like a clear decision: cutting.
I call this the knife bias. Faced with a powerful new tool, the management instinct reaches first for the blade rather than the lever. This bias is not unique to AI. (My book, chapter 6) details the mechanism in the chapter on individual fears and biases that hold back innovation, where I explain how a leader’s fear of losing their own position pushes them toward short-term trade-offs that mortgage medium-term growth.
With generative and agentic AI, the knife bias becomes expensive. Very expensive. And the numbers are starting to prove it.
The companies that cut, and are now walking it back
Ford eliminated quality engineering positions, betting on automated systems that could, in theory, catch manufacturing defects. Charles Poon, the company’s vice president of vehicle hardware engineering, eventually admitted the company had wrongly assumed that feeding AI its design requirements would produce a high-quality result without experienced human oversight. Ford brought back more than three hundred veteran engineers. CEO Jim Farley acknowledged that their return generated hundreds of millions of dollars in savings, by fixing exactly what automation alone could not handle (CNBC).
IBM ran into a narrower version of the same problem. Its AskHR assistant resolves 94% of routine HR requests. The remaining 6%, often cases requiring ethical judgment, still need a human being. IBM responded by announcing it would triple entry-level hiring across its US business in 2026. Its chief human resources officer, Nickle LaMoreaux, asked the question that should haunt every executive committee: if a company stops investing in entry-level talent, what happens in three to five years, once the pipeline runs dry (CNBC)?
Commonwealth Bank of Australia cut forty-five customer service roles, then reversed the decision in August 2025. The bank acknowledged that its initial assessment had not adequately considered all relevant business factors, which meant those roles were, in fact, not redundant (Forbes).
Klarna spent a long stretch as the poster child for replacing humans with a conversational agent. CEO Sebastian Siemiatkowski himself reversed course, explaining that it had become critical, from a brand standpoint, to assure customers that a human would always remain reachable. Klarna resumed hiring and reached its first break-even point in May 2026 (Fast Company).
Duolingo announced an “AI-first” strategy in 2025, then substantially walked it back within a year. Founder Luis von Ahn publicly admitted he had not given his teams enough context for the decision (Fast Company).
These are not isolated anecdotes. According to a Forrester survey, 55% of leaders who cut headcount citing AI now admit the decision was wrong. The firm expects roughly half of AI-attributed layoffs to be reversed in some form by the end of 2026 (Inc.com).
Notice the common thread. In every one of these cases, the company first treated AI as a subtraction tool. It then discovered, at a steep cost in recruiting, training, and lost institutional memory, that the tool’s real value sat somewhere else entirely.
What the studies say about growth
Here is the question I put to every leader I coach: have you measured what you lose in growth when you optimize only for cost?
The Boston Consulting Group studied 1,250 executives and AI leaders across nine industries. Its conclusion is stark: the companies it calls “future-built,” the 5% that have built cutting-edge AI capabilities across every function, achieve five times the revenue growth and three times the cost reduction of everyone else. These companies reinvest their gains in people and technology rather than immediately handing them back as budget savings (BCG).
IBM’s research institute reaches a structurally similar conclusion. Today, nearly half of surveyed companies’ AI spending targets efficiency. By 2030, 62% of that spending will shift toward innovation. Sixty-four percent of surveyed executives believe that, by 2030, competitive advantage will come from innovation rather than efficiency alone (IBM Newsroom).
In other words, the organizations winning with AI are not the ones cutting fastest. They are the ones redirecting the freed-up capacity toward growth: product, service, customer experience. Cost is a temporary lever, misused the moment it becomes the only objective.
How to redirect AI toward growth
My AI adoption method, detailed in (My book, chapter 14), runs in five steps: vision, communication, the group of enthusiasts I call the “SWAT team,” measured deployment, then tracking team sentiment. None of these steps begins with a headcount question. All of them begin with a direction question: what do we want the company to grow into?
Here is how to put that principle to work.
- Map value before you map costs. Identify the three processes where AI could generate new revenue, a new market, a new offer, before you even look at the roles it could automate.
- Measure what you lose in judgment. Every eliminated position carries institutional memory with it. Ask, role by role, what AI will not be able to handle in unscripted cases, exceptions, and ethical decisions.
- Route the freed-up capacity to a growth project, not just to a budget-cutting spreadsheet. Time saved through automation should fund an experiment, not only a savings line.
- Communicate before you deploy, not after. France’s postal operator La Poste defined its AI strategy in July 2023 and only told employees once it rolled out, in January 2024. The immediate result was a strike. Trust is not declared, it is measured by the alignment between what you say and what you do.
- Test, measure, adjust. An untested growth hypothesis is worth exactly as much as an untested cutting hypothesis: nothing. Kill the assumption with experimentation before it kills the project.
John, in the twelfth-floor office, ended up choosing a third path. Neither the status quo nor the nine hundred cuts. He redirected a third of the layoff budget to a team tasked with testing three new generative AI use cases in product development. Six months later, one of the three had generated a new revenue line. The other two had failed, and that was fine: together they had cost far less than nine hundred severance packages, followed a year later by nine hundred rehires.
The essentials
Remember three things:
- Cutting costs with AI is measurable within a quarter, but often regretted a year later: 55% of leaders who laid off staff for AI now admit it was a mistake.
- Companies that invest in AI for growth, rather than cost reduction alone, achieve markedly stronger results than everyone else, according to BCG.
- A structured adoption method, one that starts with vision and ends with measuring team sentiment, protects both growth and trust.
You have, like John, a whiteboard and a decision to make. AI will transform your organization, that part is already settled. What remains open is the direction you point it in: toward what you can cut, or toward what you can build.
References
- (My book) https://philippeboulanger.com/book/
- (CNBC) https://www.cnbc.com/2026/07/01/employers-who-laid-off-workers-for-ai-are-reversing-their-decisions.html
- (Forbes) https://www.forbes.com/sites/rachelwells/2026/07/26/ai-layoffs-are-backfiring-did-employers-bet-too-much-on-the-ai-boom/
- (Fast Company) https://www.fastcompany.com/91571824/the-great-ai-layoff-is-turning-into-the-great-ai-rehire
- (BCG) https://www.bcg.com/publications/2025/are-you-generating-value-from-ai-the-widening-gap
- (IBM Newsroom) https://newsroom.ibm.com/2026-01-19-ibm-study-ai-poised-to-drive-smarter-business-growth-through-2030
- (Inc.com) https://www.inc.com/bruce-crumley/55-percent-of-leaders-regret-ai-layoffs-and-a-major-hiring-reversal-has-begun/91380901
