49% of Executives Scale Back AI Agents as Usage Costs Surge

Nearly half of business executives have pulled back AI agent deployments after operating costs outpaced expected benefits, according to new research from KPMG.

The finding comes from KPMG's Global AI Pulse for Q2 2026, which surveyed 2,145 senior leaders across 20 countries at organizations with more than $50 million in annual revenue. Forty-nine percent reported scaling back AI agent initiatives specifically because costs exceeded value.

Why it matters

This isn't an AI retreat—it's a pricing reckoning. The shift from flat subscription models to usage-based token pricing has caught finance teams unprepared, and most enterprises lack the cost visibility tools to manage metered AI spending at scale. Companies that master AI cost management now will have a decisive advantage as agents become core infrastructure.

The token pricing surprise

The cost explosion stems from how AI vendors now charge for their services. Most have moved from predictable monthly subscriptions to usage-based pricing measured in tokens—small text fragments roughly equivalent to word pieces. Every input, output, and intermediate step an AI system takes consumes tokens.

AI agents amplify these costs dramatically compared to simple chatbots. Agents execute multi-step tasks, call external tools, and verify their own work. Each action triggers additional token consumption. When GitHub Copilot switched to usage-based billing on June 1, one developer projected a $180 monthly bill after a single intensive session—up from a flat $10 plan.

The visibility gap

Most enterprises are flying blind on AI spending. KPMG's companion US survey of 204 leaders at billion-dollar companies found only 26 percent have real-time visibility into AI operating costs. Globally, one-third of leaders cite limited understanding of AI cost structures, including token economics, as a barrier to agent deployment.

Yet AI investment remains strong. Seventy-nine percent of leaders call AI a top priority, up from 74 percent the previous quarter. Average AI spending held steady at $188 million. The share of organizations with AI embedded in daily operations jumped from 13 percent to 22 percent—the largest single-quarter increase KPMG has recorded.

In Asia-Pacific, 81 percent of companies report AI already delivers meaningful business value, up from 69 percent three months earlier.

Four steps to manage AI agent costs

KPMG's data points to specific actions leaders should take:

Deploy cost dashboards before scaling. Fifty-three percent of organizations globally now have AI cost monitoring in place. Real-time visibility prevents invoice-time surprises.

Educate leadership on token economics. Treat AI spending like cloud infrastructure—with owners, forecasts, and unit economics per workflow.

Embed cost review in approvals. Fifty-four percent of organizations have integrated cost-benefit analysis into AI approval processes, requiring projected ROI before any agent scales.

Rephase rather than retreat. Companies are concentrating investment where returns are strongest, not abandoning AI altogether.

The companies scaling back agents today are clearing budget for what delivers measurable value. As KPMG notes, this represents a market maturing from open-ended experimentation to financial discipline—not a bubble bursting.

These findings were first reported by Sandy Carter for Forbes, drawing on KPMG's Q2 2026 Global AI Pulse survey.