What AI agents are being used for, how often they work, and what firms plan to spend, from surveys and research papers.
of large US organizations ($1B+ revenue) are building, deploying or developing AI agents, up from 53% the quarter before (KPMG, Q3 2026)
| Group | Organizations |
|---|---|
| Building, deploying or developing agents | 62% |
| Defined no-autonomy zones for agents | 49% |
| Multi-agent systems | 25% |
Source: KPMG LLP (US), AI Quarterly Pulse Survey, Q3 2026
of large US organizations ($1B+ revenue) are building, deploying or developing AI agents, up from 53% the quarter before (KPMG, Q3 2026)
Chart studio
Switch the chart type and the style; sort it, download it or embed it. Hover a bar for the exact value.
| Group | Organizations |
|---|---|
| Building, deploying or developing agents | 62% |
| Defined no-autonomy zones for agents | 49% |
| Multi-agent systems | 25% |
Same data, any style
| Group | Scaling |
|---|---|
| Large orgs, 2025 | 27% |
| Large orgs, 2026 | 40% |
| Smaller orgs, 2026 | 22% |
| Large orgs, coding agents, 2026 | 31% |
| Group | Earlier | Latest |
|---|---|---|
| OSWorld (computer tasks) | 12% | 66% |
| Terminal-Bench (real-world tasks) | 20% | 77.3% |
Every stat
Each card links to its own page with the source, a ready citation and the data.
of large US organizations ($1B+ revenue) are building, deploying or developing AI agents, up from 53% the quarter before (KPMG, Q3 2026)
Sample: 314 US-based C-suite and business leaders at organizations with $1 billion or more in revenue, 24 Jul to 25 Aug 2026.
of large US organizations are developing or implementing multi-agent systems, up from 6% in each of the two quarters before (KPMG, Q3 2026)
Sample: 314 US-based C-suite and business leaders at organizations with $1 billion or more in revenue, 24 Jul to 25 Aug 2026.
of large US organizations have defined high-risk uses where AI agents may not make decisions on their own (KPMG, Q3 2026)
Sample: 314 US-based C-suite and business leaders at organizations with $1 billion or more in revenue, 24 Jul to 25 Aug 2026.
of large US organizations require a person to validate AI agent outputs, up from 22% in Q1 2025 (KPMG, Q1 2026)
Sample: 237 US-based C-suite and business leaders at organizations with $1 billion or more in revenue, 17 Feb to 17 Mar 2026.
of respondents at large organizations ($1B+ revenue) say they are scaling AI agents, up from 27% a year earlier; smaller organizations stayed at 22%
Sample: 1,719 online respondents in 97 nations, 4 May to 8 Jun 2026; large = more than $1 billion annual revenue (36% of respondents).
of respondents at organizations under $1 billion in revenue say they are scaling AI agents, flat on a year earlier (McKinsey 2026)
Sample: 1,719 online respondents in 97 nations, 4 May to 8 Jun 2026; large = more than $1 billion annual revenue (36% of respondents).
of respondents say their organization decided against buying software because it could be built in-house with agentic coding tools (McKinsey 2026)
Sample: 1,719 online respondents in 97 nations, 4 May to 8 Jun 2026; large = more than $1 billion annual revenue (36% of respondents).
of respondents at large organizations say they are scaling software coding agents; about two in ten overall (McKinsey 2026)
Sample: 1,719 online respondents in 97 nations, 4 May to 8 Jun 2026; large = more than $1 billion annual revenue (36% of respondents).
of McKinsey survey respondents said their organization was at least experimenting with AI agents in mid-2025
Sample: 1,993 online respondents in 105 nations, 25 Jun to 29 Jul 2025.
of McKinsey survey respondents said their organization was scaling an agentic AI system somewhere in the enterprise in mid-2025; 39% more were experimenting
Sample: 1,993 online respondents in 105 nations, 25 Jun to 29 Jul 2025.
of US senior executives said AI agents are already being adopted in their companies (PwC, May 2025)
Sample: 300 US senior executives, May 2025.
of US companies adopting AI agents say they deliver measurable value through higher productivity (PwC, May 2025)
Sample: 300 US senior executives, May 2025.
of Gartner webinar attendees said their organization had made significant investments in agentic AI; 42% had made conservative ones (Jan 2025 poll)
Sample: 3,412 Gartner webinar attendees, January 2025.
task success for AI agents on OSWorld, a test of real computer tasks, up from 12%; agents still fail about 1 in 3 attempts
success rate for AI agents on real-world tasks in Terminal-Bench, up from 20% in 2025 (Stanford AI Index 2026)
Reconciled
The different numbers you see quoted are often all real: they come from different panels, years and questions. Here they are side by side.
| Number | When | Who measured | Why it differs |
|---|---|---|---|
| 62% | Jul - Aug 2026 | KPMG AI Quarterly Pulse | 314 US leaders at $1B+ firms; counts building, deploying or developing. |
| 79% | May 2025 | PwC AI agent survey | 300 US senior executives; "being adopted" includes agent features in apps. |
| 23% scaling, 62% at least experimenting | Jun - Jul 2025 | McKinsey | Global respondents; scaling means expanding in at least one function. |
| 40% vs 22% | May - Jun 2026 | McKinsey | Scaling agents: respondents at $1B+ organizations vs smaller ones. |
| 17% | 2026 survey | Gartner CIO and Technology Executive Survey | Organizations that have deployed AI agents; a stricter test. |
Stat check
Famous stats with no real source, or quoted wrong, and where they came from.
79% of companies have adopted AI agents
PwC asked 300 US senior executives in May 2025 whether AI agents are "being adopted" in their companies, which counts agent features in everyday apps. Gartner's 2026 CIO survey puts deployed agents at 17%. Where it came from
Quote this instead: 62%40% of enterprise apps already have AI agents
A Gartner prediction (Aug 2025) that 40% of enterprise apps will include task-specific agents by the end of 2026, up from less than 5% in 2025. It is a forecast, not a measurement. Where it came from
Over 40% of AI agent projects have failed
Gartner predicted (Jun 2025) that over 40% of agentic AI projects will be canceled by the end of 2027. It is a prediction, not a count of failures. Where it came from
Every number on this page was checked at its original source. Free to cite with a link to the stat page.
Free alerts from DigitalTreed, straight to your browser.
