AI agent statistics 2026: adoption, measured.

In 2026, about two in ten organizations say they have scaled AI agents across the business, and 40% of large companies are scaling agents in at least one function. That is from McKinsey’s survey of 1,719 respondents from May to June 2026. KPMG puts agent deployment at 53% of large US companies, and Gartner forecasts $29.2 billion of spending on AI agents and assistants in 2026.

Below are the key AI agent statistics of 2026, each linked to its primary source with its date and sample. We checked every source on September 28, 2026. If you need the basics first, our guide on what an AI agent is explains the term.

01How many companies use AI agents in 2026?

It depends on what counts as use. About two in ten organizations report that AI agents have reached the scaling phase across the business, while about half of large US companies say they have deployed agents. Checked on September 28, 2026; each row links its source:

The gap between 20% and 53% is mostly about definitions and samples. KPMG asks only large US companies whether they have deployed agents at all. McKinsey asks a broader, global sample whether agents are scaling, a higher bar.

Size matters more than anything else in the data. McKinsey found large organizations scaling agents almost twice as often as smaller ones. For smaller teams, our list of agentic workflows you can run today is a more practical start than any enterprise program.

02How big is the AI agent market?

Estimates for 2026 range from $11.6 billion to $29.2 billion, a factor of 2.5, because the firms measure different things. Gartner forecasts spending on AI agents and assistants, including consumer ones, while market research firms estimate vendor revenue. Checked on September 28, 2026:

For scale, Gartner puts total worldwide AI spending at $2.67 trillion in 2026, up 49.5%, more than half of it on AI infrastructure such as servers and chips. Agents and assistants are about 1% of that.

Gartner has also published two forecasts about where agents are heading:

• In apps: 40% of enterprise applications will include task-specific AI agents by the end of 2026, up from less than 5% in 2025 (August 2025).

• In software budgets: up to $234 billion of enterprise application spending is exposed to agents doing the work across systems by 2030, about 20% of enterprise SaaS spending (July 2026).

03How many AI agent projects fail?

Nobody has measured it yet. The most quoted figure is a prediction: in June 2025, Gartner forecast that over 40% of agentic AI projects will be canceled by the end of 2027, because of rising costs, unclear business value or weak risk controls.

In the same release, Gartner estimated that only about 130 of the thousands of vendors selling “agentic AI” offer real agents. The rest, it said, rebrand chatbots, assistants and robotic process automation, a practice it calls agent washing.

What has been measured points to slow returns rather than collapse:

04How often do AI agents fail at tasks?

Often, on long and realistic work. On OSWorld 2.0, 108 desktop workflows that take a person about 1.6 hours at the median, the best agent tested, Claude Opus 4.8 at maximum thinking, completed only 20.6% of tasks. GPT-5.5 plateaued near 13%, according to the OSWorld 2.0 paper from June 2026.

The authors found that agents rarely fail at clicking or coding. They lose track of constraints, miss information that arrives mid-task, guess instead of asking the user, and skip checking their work.

Shorter, well-defined tasks go better. Stanford’s 2026 AI Index reports that agents rose from 12% to about 66% task success on the original OSWorld, yet still fail roughly one attempt in three on structured benchmarks. The best coding agent scores 58.2% on Terminal-Bench 4.0, as our coding agent leaderboard shows.

Splitting work across several agents does not remove these limits, and it multiplies the cost, which our explainer on multi-agent systems covers.

05How many developers use AI coding agents?

Almost all professional developers now use them. JetBrains found that 90% used AI coding agents at work at least weekly, and 68% daily, in a survey of more than 15,000 developers from May to July 2026.

The Stack Overflow row adds its daily, weekly and occasional users. Our AI coding statistics page covers which tools developers use, how much code agents write, and how secure that code is.

06What do people expect agents to do to jobs?

More cuts than have happened so far. In McKinsey’s 2026 survey, 39% of respondents expected AI to reduce their organization’s headcount in the coming year, up from 32%. But the share who expected cuts a year ago was about double the share who now report actual cuts.

Reported incidents are rising too. Stanford’s 2026 AI Index counted 362 documented AI incidents in 2025, up from 233 in 2024; that count covers all AI, not only agents. Our guide on whether it is safe to let an AI agent control your computer covers the controls that prevent the common failures.

07FAQ

Read next: 10 agentic workflows you can run today, AI for small business, or our log of real AI agent incidents.