KPMG’s 2026 U.S. CEO Outlook arrived with a tension that matters more than the headline optimism. As Axios reported from KPMG’s 2026 U.S. CEO Outlook, in a survey of 400 U.S. CEOs conducted Aug. 11 to Sept. 21 at companies with more than $500 million in annual revenue, 89% said they expect to increase headcount over the next three years, while 53% said they also plan workforce reductions in some areas. Because more than one-third of the companies surveyed generate over $10 billion in revenue, the results are best read as a window into large-company planning, not a census of the labor market.
For workers, managers and investors, the useful question is not whether AI will ‘create jobs’ or ‘destroy jobs.’ It is how companies can plan both expansion and cuts at once, and what proof they should provide before calling that strategy growth rather than a labor-cost program with better branding. What the survey captures is that generative AI has moved from pilot project to operating-model decision.
Why both numbers can be true
The cleanest way to read the survey is as two ledgers running at the same time.
On the growth ledger, CEOs are plainly optimistic. KPMG says 84% are confident in their own company’s growth over the next three years, 82% in their industry’s growth, 79% in U.S. economic growth and 64% in global growth. KPMG’s overview says generative AI is shaping strategy, margins and workforce design now, not waiting in the lab. That helps explain why 79% of CEOs say they plan to hire talent with AI and technology capabilities.
On the displacement ledger, the same companies are rethinking work at task level. According to Axios’s reporting on the survey, 86% are redesigning roles, career paths and team structures for human-AI collaboration. That can mean retraining and reallocation. It can also mean automating repetitive administrative work, consolidating management layers, reducing contractors, shrinking duplicated teams after acquisitions, or letting attrition take down selected functions.
Those moves are not mutually exclusive. A company can hire product managers, sales staff, AI engineers, data specialists and cybersecurity talent while reducing back-office processing work or mid-level coordination roles. It can grow total headcount and still produce painful local job losses by function, geography or job level. The survey measures executive intent, not realized employment outcomes, but it captures the operating logic clearly: AI is being treated as a reorganization tool, not just a software purchase.
What ‘role redesign’ means for the people inside the company
The phrase doing the most work in the survey is ‘redesigning roles.’ It sounds constructive, and sometimes it is. If AI removes the lowest-value tasks from a job, companies can use the time saved to deepen customer work, speed product cycles or give employees broader scopes. In that version, AI expands organizational capacity and human work shifts upward.
But the public summary does not reveal the mix between retraining, reassignment, automation and elimination. That matters because the new roles being added may not line up with the workers whose old tasks are disappearing. A firm can say it is hiring and still leave many incumbent employees worse off if the openings are concentrated in scarce, technical positions and the affected workers get little paid training or no realistic path into them.
The cyber numbers make the same point. Axios reports that 85% of CEOs expect cybersecurity threats to affect organizational prosperity over the next three years, while 48% plan to increase cybersecurity spending. That likely supports demand for security roles and controls. It does not mean the broader workforce benefits equally. AI can create new compliance, data and security work at the same moment it strips routine work from finance, support or administration.
Recent layoffs in tech show why sweeping claims are risky. Fast Company’s October roundup cites HubSpot cutting 7% of staff, Netflix reducing about 1,000 jobs, or roughly 6% of its workforce, Flock Safety trimming 18% through a voluntary program, Airtable cutting jobs after its acquisition by Bending Spoons, and Amazon Stores eliminating fewer than 1,000 roles across the U.S., India and the U.K. The reasons vary: AI-focused restructuring in some cases, post-acquisition integration in others, plus backlash and pandemic-era overhiring. That mix does not prove AI is the cause of the cuts, but it does show why ‘AI layoffs’ is too blunt a category to be useful.
The test employees and managers should demand
If companies want credit for an AI growth strategy, they should be asked to show both ledgers.
At minimum, workers, managers and investors should push for six specifics:
- baseline and target headcount by function, not just companywide totals;
- which tasks are being automated and which roles are being added;
- how much of the hiring is expansion versus replacement for attrition;
- how many employees will get paid training, redeployment support and an actual path into redesigned jobs;
- how contractor reductions, outsourcing or post-acquisition integration factor into the plan;
- what outcomes are expected beyond labor savings, including cycle time, error rates, customer satisfaction and security incidents.
Those details matter because executives have strong incentives to describe the same plan in the language of growth and efficiency. Buying AI licenses, flattening teams and holding down labor costs may improve margins. That is not the same thing as expanding the business’s capacity in a way that creates durable work. A company making the stronger bet will be able to show that AI spending is tied to more revenue-producing activity, better service or stronger security, and that a meaningful share of current employees can move into the new model.
The KPMG survey is valuable precisely because it makes the contradiction visible instead of hiding it. CEOs are not, at least in this snapshot, choosing between hiring and cutting. Many are planning to do both. For people whose jobs sit closest to repetitive workflows, that means executive optimism is not protection. For managers trying to build credible staffing plans, it means the right question is no longer ‘Are we adopting AI?’ but ‘Which work grows, which work disappears, and who gets carried into the redesign?’ Until companies answer that with numbers, ‘more hiring’ is a plan, not yet proof.
By
By
By
By
By
