Stop Firing Your Future: Why AI-Attributed Layoffs Are the Most Expensive Mistake in Enterprise Transformation

Something extraordinary is happening in corporate America right now, and very few people are saying what needs to be said about it.

In 2025, U.S. companies announced 1.2 million job cuts, nearly twice the prior year's total. Approximately 55,000 of those cuts explicitly cited artificial intelligence as the reason. In the first quarter of 2026, the pace accelerated: more than 20% of confirmed tech layoffs worldwide now explicitly reference AI and automation, up from under 8% just a year ago. Block cut 40% of its workforce. Amazon eliminated 30,000 corporate roles. Meta, Microsoft, Salesforce, Workday, CrowdStrike, and dozens of others followed. CEOs are no longer euphemistic about it. They are openly attributing headcount reductions to AI capabilities.1

The market has applauded. Investors reward the announcements. Boards ask for more. The narrative has taken hold: AI is here, it can do the work, and the organizations that move fastest to reduce headcount will win.

The data tells a completely different story.

The layoffs are not based on proven AI capability. They are based on anticipated capability. In a survey of more than 1,000 global executives published by Harvard Business Review, the vast majority of AI-attributed layoffs were found to be anticipatory: based on what companies expect AI to deliver, not what it has already delivered. A separate study put a finer point on it: 60% of executives said they made headcount reductions in anticipation of AI efficiencies. Only 2% said they had made large staff reductions as a result of actual AI implementation.2

Read that again. Two percent.

Oxford Economics applied a simple economic test: if machines were truly replacing humans at scale, output per remaining worker should be skyrocketing. The productivity data does not support that conclusion. Their assessment was blunt: some firms are dressing up routine cost-cutting as innovation, and the primary motivation appears to be investor relations. Meanwhile, only 16% of individual workers had high AI readiness in 2025, and only 23% of organizations offered any kind of AI training. Companies are making permanent workforce decisions based on a promise, cutting people for not being productive with tools their employers never trained them to use.3

If these were sound strategic decisions, the organizations making them would be reporting improved performance and moving on. Instead, a different pattern is emerging: regret, followed by quiet reversal.

According to Forrester, 55% of employers who made AI-attributed cuts in 2025 already regret the decision, and half are expected to quietly rehire.4 Gartner reinforces this with a specific prediction: by 2027, 50% of companies that attributed headcount reduction to AI will rehire staff to perform similar functions, often under different job titles. And Gartner's survey of 321 customer service leaders revealed that only 20% had actually reduced staffing because of AI. The vast majority of recent workforce reductions were influenced by broader economic conditions rather than automation. These are not companies replacing workers with technology. These are companies cutting costs and calling it innovation.5

The most visible example is Klarna. The Swedish fintech's CEO declared that AI could do all of the jobs humans do, reduced headcount from 5,500 to 3,400 according to the company's own IPO prospectus, and celebrated $10 million in annual marketing savings alone. Then customer satisfaction deteriorated as the CEO himself later acknowledged, telling Bloomberg that cost had been "a too predominant evaluation factor" and that "what you end up having is lower quality." By late 2025, the CEO reversed course, publicly stating that "really investing in the quality of the human support is the way of the future for us" and launching a new initiative to hire customers as human support agents. The company that was supposed to prove AI replaces human labor became the canonical cautionary tale of the 2025-2026 wave.6

Klarna is not an outlier. The pattern it established (aggressive efficiency claim, followed by quality degradation, followed by strategic reversal) has been documented across enough organizations that analysts now treat it as a structural expectation rather than an exception. The institutional knowledge left with the people. The AI systems were deployed without the human expertise needed to direct them effectively. And the output quality gap that followed created the rehire pressure that the data now documents across the 2025-2026 wave.

The financial calculus behind AI-attributed layoffs typically models labor cost savings. It rarely models what those people took with them when they left.

Gartner puts it directly: AI is not mature enough to fully replace the expertise, empathy, and judgment that human workers provide. Relying solely on AI right now is premature and could lead to unintended consequences. Organizations that cut too fast and too deep are discovering that the technology requires more human oversight than the initial business case anticipated, not less.5 The AI did not replace the human judgment. It required more of it.

Here is what the financial models miss. Every Fortune 500 organization runs on institutional knowledge that does not exist in any system, any process document, or any training dataset. It exists in the heads of experienced employees who understand why the process works the way it does, what the exceptions are, which customer relationships require special handling, what the undocumented dependencies are between systems, and what judgment calls get made fifty times a day that no one ever wrote down. McKinsey's research confirms what practitioners already know: most enterprise processes remain exactly this, tacit knowledge locked in the heads of experienced employees rather than documented in any system.7

When those employees leave, that knowledge leaves with them permanently. You cannot fine-tune an AI model to recover it. You cannot hire replacements who have it. You cannot consult your way back to it. It is gone.

And here is the connection that almost no one is making: that tacit knowledge is precisely what you need to redesign your workflows for AI. The people who understand how the business actually operates (not how the process documentation says it operates, but how it actually works) are the people who must be in the room when the organization redesigns how work should be done in an AI-enabled world. Firing them before the redesign is complete is like burning the blueprints before construction starts.

To understand why this is a strategic error, consider what a well-executed AI transformation actually requires. The research across every major consulting firm and academic institution studying this issue converges on the same sequence:

First, understand what AI makes possible. Then set enterprise strategy informed by that understanding. Then decompose that strategy into specific Business Transformation Imperatives. Then educate the workforce at every level on what AI can do. Then redesign workflows for how work should actually be done with humans and AI working together. Then redesign jobs based on the redesigned workflows. Then deploy AI tools that enable the new workflows. Then, and only then, right-size the organization to fit the new operating model.

The data on where organizations actually stand is sobering. Only 6% qualify as AI high performers according to McKinsey.7 Only 5% are achieving AI value at scale according to BCG.8 Only 21% have fundamentally redesigned any workflows.7 Only 34% are truly reimagining their business rather than layering AI onto existing processes.9 The vast majority of Fortune 500 companies have not even begun the transformation work that these layoffs presuppose.

What these companies are doing is right-sizing first, before the transformation is designed, before the workflows are redesigned, before the jobs are redefined, and before the AI capabilities are deployed in production and proven at scale. They are demolishing the bridge before they have built the new one.

Every person in a Fortune 500 company today is needed for one of two things: running the business in its current form (which does not stop during transformation), or helping transform the business into its future form. Both of these require people who understand the business. Cutting headcount before the transformation is complete means you either cannot maintain current operations, or you cannot execute the transformation. In most cases, both.

The strategic error is not limited to the people who leave. It extends to the people who stay.

Research identifies a growing segment of employees who have simply stopped trying, workers who do not believe their employer deserves their discretionary effort. This group continues to grow, and the trend shows no signs of reversing. The cause is not mysterious. Employees watch colleagues get fired for AI that does not work. They are told to be grateful they still have jobs. They are asked to absorb the work of eliminated positions while also embracing the AI tools that their employer just used to justify firing their colleagues. The result is a workforce that has stopped extending itself.

This is precisely the opposite of what AI transformation requires. McKinsey's research shows that companies involving at least 7% of employees in transformation initiatives double their chances of delivering positive results, with the highest performers involving 21% to 30%.7 The transformation that these organizations need to execute is the most complex operational change in the history of business. It requires creativity, collaboration, domain expertise, and discretionary effort from every level of the organization. It requires people who are willing to reimagine how their own work should be done, people who trust their employer enough to engage honestly with a process that will fundamentally change their roles.

A workforce that has watched its colleagues get fired for AI that does not yet work is not a workforce that will extend that trust. No amount of AI compensates for the productivity loss of a demoralized, disengaged organization.

The organizations that will emerge from this era as market leaders are not the ones cutting headcount fastest. They are the ones investing in their people and building the organizational capability to transform.

This means keeping the people who understand your business. Every person who understands your domain, your customers, your operations, your exceptions, your dependencies is a person you need for the transformation ahead. Rather than cutting headcount, redeploy people toward the transformation effort itself. The same employees whose current roles may eventually be redesigned are the employees who understand those roles well enough to help redesign them.

It means building AI fluency at every level of the organization. Not a briefing deck. Not a vendor demo. Substantive, working engagements where employees at every level experience what AI makes possible and develop enough understanding to make informed decisions about how their own work should change. BCG found that persona-based learning journeys deliver AI adoption at a level 20 times higher than a broad-based approach.10 The investment in education pays for itself in transformation quality.

It means redesigning workflows before deploying AI tools. Organizations that take a "work-backward" approach, redesigning work first and then deploying technology that enables the new design, achieve dramatically better outcomes than those taking a "tech-forward" approach. One global financial services firm that deconstructed tasks before deploying AI achieved a 59% workload reduction according to MIT Sloan Management Review, precisely because the redesign came first.11

It means redesigning jobs based on the redesigned workflows, not based on assumptions about what AI might eventually be able to do. When the workflows are redesigned, the new role requirements become clear: what skills are needed, what judgment calls humans make, what oversight AI requires, what new capabilities the organization needs to build. The right-sizing decisions become evidence-based rather than anticipatory.

And it means right-sizing only after the transformation is implemented and proven, when you have real data on what the new operating model requires rather than projections based on promises.

This is not slower. It is dramatically faster than the alternative, because the alternative (cut first, transform later) produces the regret cycle, the rehire cycle, the institutional knowledge loss, and the morale collapse that the data now documents. The methodology we have developed, detailed across this article series, tells you how to execute each step. And it starts from a fundamentally different premise: that the way you succeed with AI is by investing in your people, not by eliminating them.

We should be honest about this. There are situations where headcount adjustments during transformation are appropriate. When a specific AI capability has been deployed in production, proven at scale, and demonstrated that it can reliably perform the work, and when the person's skills cannot be redeployed to the transformation effort or other value-creating roles, then a headcount adjustment is a legitimate business decision.

But the bar should be high. The questions to ask before any AI-attributed workforce reduction: Has the AI system been deployed in production (not in pilot)? Has it been proven at scale (not in a controlled test)? What are its error rates, and how does it handle edge cases? What is the rollback plan if it fails? How will institutional knowledge be retained? Has the workflow been redesigned, or are we automating the old process? Can this person be redeployed to the transformation effort itself? Are we going to grow top line revenue via new products or services or new markets or customer segments? If so, can I use these people to help with that growth?

If the answers to these questions are not grounded in evidence, the layoff is premature. The data is unambiguous on this point.

Two years from now, two very different types of organizations will be competing in the same markets.

The first type will have cut headcount aggressively, lost institutional knowledge, demoralized their remaining workforce, struggled to execute transformation without the domain expertise they eliminated, and spent significant resources rehiring (often offshore, often at lower quality) to fill the gaps they created. They will have saved money on labor costs in the short term and lost competitive position in the medium term.

The second type will have kept their people, invested in AI fluency at every level, redesigned workflows with the employees who understand them, redeployed people into roles defined by the new operating model, and built an organization that transforms continuously rather than episodically. They will have spent more on workforce investment in the short term and built a compounding competitive advantage that their leaner competitors cannot replicate.

The competitive advantage in AI transformation is not the technology. Everyone has access to the same models, the same platforms, the same vendor ecosystems. The competitive advantage is organizational: the domain expertise, the institutional knowledge, the workforce capability to transform, and the trust between employer and employee that makes transformation possible.

The companies that are destroying that advantage in the name of AI efficiency are making the most expensive mistake in enterprise transformation. The data already proves it. The question is whether more organizations will make the same mistake before the evidence becomes impossible to ignore.

This article is the first in a 27-article series on Plaster Group’s AI Business Transformation Methodology. The series provides a comprehensive, research-backed methodology for executing AI transformation at every level of the organization, from strategy through continuous optimization, without sacrificing the workforce capability that makes transformation possible.