KPMG AI Quarterly Pulse Survey, Q3 2026
TL;DR
KPMG US’s quarterly survey of large-company leaders, released 24 September 2026. 314 US C-suite and business leaders at organisations with $1bn+ revenue, surveyed 24 July to 25 August 2026.
- 58% report measurable value. “Nearly 6 in 10 leaders (58%) report measurable business value from their AI initiatives” (deck p. 2). Productivity is the most common kind (55%); financial outcomes 37%.
- Agents, by one count, 62%. The release: “62% of organizations report they are now building, deploying or developing AI agents, up from 53% last quarter.” The deck labels the same 62% “Deployment” and compares it with 42% in Q3 2025 (p. 3).
- Multi-agent work jumps. Organisations actively developing or implementing multi-agent systems: 25%, against 6% in each of the two previous quarters (release; deck p. 3).
- Governance confidence jumps too. 73% are confident in their governance and capabilities to manage AI risk at scale, up from 57% last quarter; 49% have defined high-risk cases where agents may not decide alone (deck p. 5).
- Cost discipline. 74% include cost reviews in AI approvals (61% last quarter), 70% use monitoring dashboards, 43% have usage or token budgets (deck p. 2).
Key data
| Measure | Q3 2026 | Comparison | Page |
|---|---|---|---|
| Measurable business value from AI | 58% | none given | deck 2 |
| Value in productivity / decision-making / financial outcomes | 55% / 49% / 37% | none given | deck 2 |
| Agents: “Deployment” (deck); “building, deploying or developing” (release) | 62% | 42% Q3 2025 (deck); 53% Q2 2026 (release) | deck 3 |
| Piloting AI agents | 25% | 55% Q3 2025 | deck 3 |
| Developing or implementing multi-agent systems | 25% | 6% Q2 2026 | deck 3 |
| Data readiness and access as top barrier to agents | 53% | none given | deck 3 |
| Significant workforce adoption | 44% | 23% last quarter; 10% a year ago | deck 4 |
| Confident in governance to manage AI risk at scale | 73% | 57% last quarter | deck 5 |
| High-risk cases where agent autonomy is not allowed | 49% | 28% Q3 2025 | deck 5 |
| Projected AI investment, next 12 months | $173M | $130M Q3 2025 | deck 2 |
Method and definitions
- Sample: 314 US-based C-suite and business leaders at organisations with annual revenue of $1bn or more; more than a third have revenue of $10bn or more. Fieldwork 24 July to 25 August 2026. Part of KPMG’s Global AI Pulse.
- “Measurable business value” is the respondent’s own judgement. No threshold, metric or question wording is given.
- The sample changes each quarter. Q2 2026 had 204 respondents (28 April to 25 May 2026); Q3 has 314.
How the agent series is worded across quarters. The Dutch report says KPMG changes its agent definitions between quarters. The releases show changes in wording; they do not document a change of question.
| Quarter | Headline | Wording |
|---|---|---|
| Q1 2025 | 11% | “full deployment remains stagnant at 11%“ |
| Q2 2025 | 33% | “deployed at least some agents” |
| Q3 2025 | 42% | “Deployment” (in the Q3 2026 deck) |
| Q1 2026 | 55% | “AI agent deployment” (in the Q2 2026 deck) |
| Q2 2026 | 53% | “AI agent deployment holds steady” |
| Q3 2026 | 62% | “building, deploying or developing AI agents” (release); “Deployment” (deck) |
The Q2 2026 deck (p. 6) shows the headline next to sub-items. Three of them (scaling agents across functions 29%, multi-agent systems 6%, orchestrating agents across workflows 18%) add up to the Q2 headline of 53%. The same three add up to 48% for Q1, against a stated 55%. How the headline is composed is not documented. The investment series also jumps: Q2 2026 reports a weighted average of $202M (Q1: $207M), Q3 reports $173M and compares it with Q3 2025, not with Q2.
Scope and reliability
Self-report by senior leaders at large US companies, by a firm that sells AI advisory services. Quarterly samples of 200-300 change in size and are not described as a panel. The headline wording for agents widens over time, from “full deployment” to “at least some” to “building, deploying or developing”. Read the agent figures as a series of separate snapshots, not as a clean trend. The value question has no threshold, which puts it among the broad measures.
Dynamic-capabilities reading
digital-transforming/improving-digital-maturity: governance confidence, cost reviews, monitoring dashboards and token budgets are reported as signs of maturing AI operations.digital-seizing/balancing-digital-portfolios: the move from piloting (55% to 25%) to deployment, and cost reviews in AI approval, are portfolio decisions.contextual/internal-barriers: data readiness and access is the top barrier to agent deployment (53%).
How it connects
- enterprise-ai-adoption and the thread ai-maturity-measurement-comparison: a broad measure of value (58%), on a US large-firm sample.
- PwC Global CEO Survey 2026: a strict measure from CEOs in 95 countries, 12% reporting both cost and revenue benefits. The two differ in threshold, respondents, population and date.
- Deloitte State of AI in the Enterprise 2026: 21% report a mature model for governing agents, a different question from KPMG’s governance confidence.
- Gillespie et al. (KPMG): a different KPMG instrument, surveying workers in 47 countries rather than US executives.
- ai-agents and agent-oversight-and-delegation: multi-agent systems, and the 49% that bar agent autonomy in high-risk cases.
- Wharton and GBK, 2025 AI Adoption Report: another broad measure among large US firms: 74% report positive ROI.
What was actually ingested
The Q3 2026 deck (6 pages, from KPMG’s own asset URL) and the Q3 release, both in full. For the series: the Q2 2026 deck and release (saved to raw), and the Q1 and Q2 2025 releases (quoted in the raw file’s notes, not saved). Chart values were read from page renders, because pdftotext scrambles them. The industry decks (banking, technology, asset management and private equity) were not read.
Linked entities and concepts
- KPMG — publisher (KPMG US).
- KPMG: publisher, no entity page yet. With Gillespie et al. this is the second KPMG-published source (see the batch’s entity proposals).
- enterprise-ai-adoption, ai-agents, agent-oversight-and-delegation: concepts.
Dangling (single-source mention, deferred): Todd Lohr (KPMG LLP, quoted), Rahsaan Shears (AI Enterprise Transformation Leader, KPMG LLP, quoted in Q2 and Q3 releases).