Is agentic AI paying off or just hype?

A vendor-run survey says 96 percent of organizations with agentic AI in production met or beat their ROI expectations. McKinsey finds 37 percent of respondents attribute any profit effect to AI, and Gartner forecasts over 40 percent of agentic projects canceled by 2027. Read 2026-11-06.

By Mackenzie Bowes · mackenzie@mackenziebowes.com · @_mackenziebowes on X

Contents

The same technology gets opposite verdicts in 2026 because the sources count different things. A survey commissioned with a voice-AI vendor reports 96 percent meeting ROI expectations. Gartner forecasts that over 40 percent of agentic AI projects will be canceled by the end of 2027, and places the field at the "Peak of Inflated Expectations". McKinsey's survey asks about profit, not expectations. This page lists what each source counted, who ran it and what it leaves out. It is a companion to AI pilot purgatory, which does the same for the figures on pilots that never ship.

Who was asked about the 96 percent that met or exceeded

CCW Digital and SoundHound AI, a company that sells voice AI, published the survey on 18 June 2026. Its release says "96% of organizations with active agentic AI deployments report results that met (54%) or exceeded (42%) their ROI expectations."[1] Three details decide what that number means:

What McKinsey asked 1,719 people about profit

McKinsey's State of AI survey was "in the field from May 4 to June 8, 2026, and garnered responses from 1,719 participants in 97 nations".[2] It reports that 37 percent of respondents "attribute at least some EBIT impact to AI use (about the same share as last year)", and that AI high performers, who attribute at least 5 percent of EBIT to AI, are "about 6 percent of all respondents".[2] On agents: "Forty percent of respondents from large organizations ... report scaling AI agents, up from 27 percent last year", while smaller organizations stayed "flat at 22 percent".[2] About 20 percent report that AI operating costs, including tokens, constrained their AI use.[2] The same survey finds that 80 percent report AI improved their individual productivity.[2] These cover AI in general; only the scaling and coding-agent figures are specific to agents.

Gartner gave a forecast and a position on the curve

Gartner's June 2025 press release predicts "Over 40% of agentic AI projects will be canceled by the end of 2027, due to escalating costs, unclear business value or inadequate risk controls."[3] That is a forecast, not a count. Its analyst says "Most agentic AI propositions lack significant value or return on investment (ROI)", and that many products are rebranded: "Gartner estimates only about 130 of the thousands of agentic AI vendors are real."[3] The release rests on a January 2025 poll of 3,412 webinar attendees, of whom 19 percent reported significant investment in agentic AI and 42 percent conservative investment.[3]

Gartner's 2026 Hype Cycle article says "Agentic AI sits at the Peak of Inflated Expectations", citing its CIO survey: "only 17% of organizations have deployed AI agents to date, yet more than 60% expect to do so within the next two years."[4] It adds that "most deployments remain narrowly scoped, and fully autonomous agents are not ready for the majority of enterprise use cases."[4] Gartner sells the research and the Hype Cycle is its own scheme, so the label is Gartner's reading.

What an agent can do on work tasks

TheAgentCompany benchmark (Carnegie Mellon and collaborators) tested agents on tasks in a simulated small software company. It reports that "the most competitive agent can complete 30% of tasks autonomously", with "a good portion of simpler tasks" solved and "more difficult long-horizon tasks" beyond reach.[5] The paper was submitted in December 2024 and last revised in September 2025, so it tests the models of that time and is a preprint. Do AI agents actually work? sets this benchmark beside three others.

How they fit together

The sources do not contradict each other as much as the headlines do. The 96 percent is a share of organizations that got to production, judged against their own expectations. McKinsey's 37 percent is a share of all respondents who see a profit effect. Gartner's 40 percent is a forecast about projects, and its 17 percent is deployment. None counts the same thing, and none reports a return in money for a typical small business.

Not yet proven

If you run an AI agent in your business and have tracked what it saved against what it cost, I would like to compare notes: mackenzie@mackenziebowes.com.

Sources

  1. SoundHound AI with CCW Digital, Research Finds 96% of Organizations Report that Agentic AI Deployments Met or Exceeded ROI Expectations in 2026, 18 June 2026. A vendor's release. Read in Rotunda 2026-11-06. www.soundhound.com/newsroom/research-finds-96-of-organizations-report-that-agentic-ai-deployments-met-or-exceeded-roi-expectations-in-2026
  2. McKinsey, The State of AI: Global Survey 2026. Read in Rotunda 2026-11-06. www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai
  3. Gartner, Gartner Predicts Over 40% of Agentic AI Projects Will Be Canceled by End of 2027, press release dated 2025-06-25. Read in Rotunda 2026-11-06. www.gartner.com/en/newsroom/press-releases/2025-06-25-gartner-predicts-over-40-percent-of-agentic-ai-projects-will-be-canceled-by-end-of-2027
  4. Gartner, 2026 Hype Cycle for Agentic AI. Read in Rotunda 2026-11-06. www.gartner.com/en/articles/hype-cycle-for-agentic-ai
  5. Xu et al., TheAgentCompany: Benchmarking LLM Agents on Consequential Real World Tasks, arXiv 2412.14161 v3, 10 September 2025 (abstract). Read in Rotunda 2026-11-06. arxiv.org/abs/2412.14161

Terms on this page

  • Agentic AI: AI that works toward a goal by planning steps, making decisions and acting with more or less independence.
  • AI agent: A system where a language model decides for itself how to do a task and which tools to use.