A new Gartner survey cuts against the simplest AI business story: replace workers, lower costs, show returns. Gartner says that pattern is happening, but it is not what separates stronger AI adopters from weaker ones.
In a May 5 newsroom release, Gartner said roughly 80% of large organizations piloting or deploying autonomous business capabilities reported workforce reductions. But the firm also said job-cut rates were nearly equal among companies reporting higher returns and those seeing modest or negative outcomes.
The cleaner takeaway is uncomfortable for executives. AI layoffs can create budget room. They do not automatically create return on investment.
The finding executives should not ignore
Gartner surveyed 350 global business executives in Q3 2025 at organizations with at least $1 billion in annual revenue or equivalent. Respondents were already piloting or deploying technologies such as AI agents, intelligent automation, or autonomous systems.
That makes the data especially relevant to early enterprise adopters. These are not companies asking whether AI agents are real. They are already spending, restructuring, and looking for results.
Gartner distinguished VP analyst Helen Poitevin put the point bluntly: workforce reductions may create budget room, but they do not create return. The companies doing better are investing in skills, roles, and operating models that let people guide and scale autonomous systems.
Computerworld, citing Gartner's analysis, reported the same pattern in ROI terms: there was no clear correlation between AI-related layoffs and stronger returns. It also noted that some organizations had reduced headcount by up to 20%, but the better-performing companies were measuring broader value, including productivity, faster time to market, revenue growth, and workforce transformation.
The agent boom is still real
This does not mean the AI-agent market is slowing. Gartner forecasts AI agent software spending will rise from $86.4 billion in 2025 to $206.5 billion in 2026 and $376.3 billion in 2027.
That is the tension. Companies are preparing for a massive shift in software budgets while still trying to prove the economics. If the first board-level metric is headcount reduction, many teams may optimize for the easiest number to show rather than the most durable value to build.
For software buyers, this matters because agent rollouts are not just procurement decisions. They change approval chains, escalation paths, quality checks, security reviews, and how teams decide whether work is good enough to ship. Cutting people before those systems are redesigned can remove the very expertise needed to make automation useful.
What stronger adopters appear to do differently
The higher-return story is less dramatic than the layoff story: retrain teams, define new roles, change operating models, and give humans clearer responsibility for AI oversight.
That means companies may need fewer people doing some repetitive work, but more people handling governance, exception management, process design, model evaluation, customer trust, and internal enablement. The work shifts rather than simply disappears.
IT Brief Australia highlighted the same point from Gartner's release: organizations with better returns were more likely to invest in the skills and structures needed to manage autonomous systems, not just reduce payroll.
The practical risk is that companies treat agents like a cost-cutting plug-in. Agents work best when they are embedded into a workflow with clear limits, human review, and a business target beyond fewer seats. Without that, the savings can be real but shallow.
The next metric to watch
The next useful signal will not be how many companies announce AI-related cuts. It will be how many can show revenue growth, faster cycle times, better customer outcomes, or lower operational risk after deploying agents.
Gartner's message is not that AI will protect every job. It is that layoffs are a weak proxy for AI success. If agent software spending more than quadruples from 2025 to 2027, companies will need a better scorecard than headcount math.
