AI-driven job cuts could lead to costly rehiring

AI-driven staffing cuts carry a rehiring risk

An executive who cuts jobs assuming AI can do the work may later pay more to restore the same human capacity. Oracle, Amazon and Salesforce have reduced their workforces in preparation for greater automation: using technology to carry out work with less human intervention. These decisions can reflect expectations about technology or a change in business direction, so executives considering similar cuts need to understand what work AI has actually taken over.

Getting that decision wrong could be costly. Consulting firm Gartner forecasts that 30% of employees dismissed because AI replaced them will need to be rehired by 2029, likely at a substantially higher cost. Gartner sells research and advisory services, so demand for help with AI decisions benefits its business. Its forecast points to a staffing risk: the anticipated savings depend on the replacement working well enough to sustain the business.

Layoff explanations describe different business decisions

To understand that replacement risk, Gartner examined employers’ reasons for reducing staff. A team led by Tori Paulman, VP Analyst at Gartner, studied more than 1 million layoffs in 2025. Less than 1% of all layoffs analyzed were attributed to AI making individual employees productive enough that their employers needed fewer people. That percentage covers all the layoffs studied, a broader group than cuts publicly associated with AI.

Productivity-driven reductions require a specific chain of events: AI increases an employee’s output, the increase reduces the number of employees needed, and the employer then cuts staffing. Gartner found that employers rarely gave this explanation for the layoffs its team examined. Other AI-attributed layoffs, meaning cuts publicly associated with AI, can follow a different business decision, such as leaving an unprofitable market or moving investment into new products.

Gartner examined one such decision separately, classifying 17% of AI-attributed layoffs in the first half of 2025 as commercial pivots. These employers shifted away from unprofitable business units toward AI product lines. Because this percentage covers AI-attributed layoffs during a six-month period, it describes a different group from the less-than-1% finding. In a commercial pivot, a company eliminates work because it is leaving a business, while replacement requires AI to perform work the company still needs.

The difference also matters when executives read an employer’s explanation. Paulman described some explanations as possible “AI-washing,” in which an AI narrative accompanies what may be conventional restructuring. She told Channel Dive that Gartner’s interest grew when layoff headlines appeared to come from technology companies that also sold AI products. Those sellers benefit from confidence in AI, and her team questioned whether their staffing announcements were intended to reinforce their productivity claims.

The timing of announcements matters too. Paulman said AI-related firings declined in the second half of 2025, even though prominent announcements could give executives an impression of accelerating replacement. Staffing plans extending through 2029 also need to account for capabilities changing after those observations. A system that becomes capable of handling more work may support a different decision, but that decision still depends on its performance at the time.

For current decisions, Paulman focuses on how much work systems can assume. Agentic AI consists of systems that can take actions across a task with some autonomy. “Agentic AI is not taking over a significant enough portion of work for an organization of any size to be successful with laying off a considerable amount of workers,” she told Channel Dive. Her concern is how much of an organization’s work these systems can handle successfully.

Replacement reversals show the financial stakes. Klarna replaced 700 customer service agents with AI in 2024, then began hiring people back by early 2026. CNBC reported that Ford, IBM and Commonwealth Bank of Australia also reversed AI-related layoffs, with engineers, human resources representatives and customer service staff collectively among the roles brought back. These employers returned to hiring after reducing human capacity, making the cost and availability of replacement staff part of the automation decision.

Productivity gains have to become business value

The reversals raise a question that comes before any staffing reduction: does AI improve the business enough to support the change? Paulman argues that everyday AI tools are failing to deliver the scale of productivity improvement many executives expect, an assessment she says is reflected across Gartner and external research. Even when a tool improves an individual task, the staffing decision depends on that task’s importance, the value created and the human work that remains.

A task’s importance depends partly on its place in the organization’s core operations. A study by software company Aptean involving more than 1,500 decision-makers found that fewer than half of organizations considered AI essential to their core work. As a software seller, Aptean has a commercial interest in organizations’ technology investment decisions. Its finding concerns AI’s role in essential activities, where improvements are more likely to affect the capacity a business needs.

Even in important activities, higher productivity has to produce a useful business result. Accenture’s Pulse of Change report surveyed 6,000 C-suite executives, senior company leaders, and non-C-suite employees; among respondents experiencing positive productivity gains, fewer than a quarter reported business value they could point to. Accenture sells consulting and technology services, including AI services, so spending on AI implementation benefits its business. Its finding concerns the conversion of productivity into value, while Aptean examines AI’s place in core work and Gartner examines employers’ reasons for layoffs.

Turning tool use into value also requires employees to adopt the technology effectively. Paulman says major technology companies face the same problems as their customers: limited employee adoption, uncertain productivity value and unrealistic executive expectations. Selling AI gives a company no automatic advantage in getting its own employees to use it well. An executive mandate can therefore reach an entire organization without delivering the expected improvements.

Paulman links that gap partly to leaders’ understanding of the technology they are deploying. Many executives working with Gartner face pressure to reduce costs and increase revenue while lacking basic knowledge of AI, data, analytics (using information to understand results) and machine learning (training computer systems to identify patterns in data). “Executives are being steeped in an environment of what I would call magical thinking: that AI will just figure it out,” she said. Treating AI as interchangeable with people can then lead leaders to choose tools that do not fit the problems they need to solve.

Replacement decisions change customer and workforce relationships

A poor fit becomes especially costly when it changes the service customers receive. Customer service has been a particular problem for some employers because the work combines task completion with interaction, context and judgment. A system may complete parts of a request successfully while leaving customers dissatisfied with the exchange. Customer acceptance therefore becomes a business requirement alongside technical performance.

Voice interactions show why acceptance matters. “Customers just don’t really like talking with AI, particularly over voice,” Paulman said. If customers resist the interaction, completing their requests may still leave the employer with a service problem. Employees’ relationships with customers and knowledge of earlier conversations can remain valuable even when AI handles some service tasks.

Preserving those relationships becomes harder once an employer has removed the people who maintain them. Paulman links baby boomer retirements and falling birth rates to a declining global labor supply, which can increase competition when companies seek to hire again. If AI replacement proves inadequate or customers reject the experience, the employer may return to the labor market under less favorable conditions. Competition for staff is one reason restoring capacity could cost more.

Employees are also responding to the possibility of replacement. In a survey attributed to research firm Gallup, frequent AI users were more than twice as likely to fear losing their jobs within five years as people who used AI only a few times a month or year. The finding describes an association between use and fear; it does not establish that one causes the other. Familiarity with AI can therefore coexist with substantial employment anxiety.

Economic growth could coexist with employment losses as well. Anthropic describes scenarios combining a 15% expansion of the U.S. economy by 2030 with unemployment reaching 18% among many white-collar workers. Anthropic develops AI systems and benefits commercially from AI adoption; these figures describe its scenarios for future economic outcomes. For an individual employer, growth in the wider economy still leaves the question of which people and capabilities its own business requires.

Assess the work and preserve human capabilities

Answering that question requires executives to establish that AI can perform the relevant work to the required standard before eliminating a role. They also need to connect the productivity improvement to meaningful business value, including the quality of service customers receive. Success at an individual task is one input to this assessment. The role may also carry responsibilities for judgment, relationships and institutional knowledge that remain unchanged.

Gartner recommends treating AI as a complement to human capabilities, using the term “toolmate” for that relationship. Its advice is to resist indiscriminate automation and the delegation of every decision to AI. For workforce planning, this means identifying where technology improves performance while retaining the human context and knowledge the organization needs. Gartner’s intended outcome is stronger human performance with fewer unnecessary cycles of dismissal and rehiring.

Choosing where to use AI requires leaders to distinguish between technical approaches. Paulman argues that they should understand when a problem calls for large language models, or LLMs, which generate language, and when it calls for automation that carries out defined processes. Starting with the work makes that choice more precise and helps identify where people remain necessary. Assuming that an AI system is equivalent to a person obscures both decisions.

The choice of technology also creates responsibilities for everyone involved in its use. “It is the job of the user of the tool, the builder of the models and also the executive to use this technology in a verifiable, correct way,” Paulman said. She also emphasizes usefulness, giving leaders a business requirement alongside technical correctness. A deployment needs to produce a result that people can check and that serves a meaningful purpose.

People need clear explanations to judge whether a deployment meets those requirements. Paulman urges information technology (IT) professionals to explain AI plainly to community members, parents and children as well as colleagues, because the transition affects how people understand human work. Within a business, that means explaining what a system does, how its results can be checked and where human responsibility remains. Clear explanations can help people make more informed decisions and reduce fear.

Key executive takeaways

• Verify what drove the cuts: Gartner forecasts that 30% of employees dismissed because AI replaced them will need to be rehired by 2029, likely at higher cost. Executives assessing peers’ layoffs need to distinguish proven automation gains from shifts away from unprofitable businesses.
• Connect productivity to business value: Among Accenture survey respondents reporting productivity gains, fewer than a quarter reported identifiable business value. Business owners evaluating staffing reductions need evidence that AI improves core operations and sustains the required quality of work.
• Include customers and rehiring in the calculation: Customer resistance to AI service and the loss of employee knowledge can undermine expected savings. Workforce and service leaders need to assess customer acceptance, remaining human responsibilities and the cost of restoring capacity before cutting roles.
• Match the technology to the work: Gartner recommends using AI to strengthen human performance while preserving judgment, relationships and institutional knowledge. Technology and business leaders need to define the task, select the appropriate tool and establish how people will verify its results before eliminating roles.