Ramesh Raskar's team at MIT's NANDA initiative put a number on enterprise AI's most uncomfortable secret in July 2025, and the number has held up since1. Surveys through 2026 confirmed it, refined it, or found new ways to describe the same gap56. 95% of firms running generative AI pilots reached zero measurable profit impact. That number came from 300-plus disclosed projects, 52 interviews and a survey of 153 leaders. Gartner measured the forward half of the same story2. More than 40% of agentic AI projects now running inside companies will be canceled by the end of 2027, according to Senior Director Analyst Anushree Verma. Two research houses, working apart, reached one conclusion. Enthusiasm ran past proof.
Generative AI pilots left profit flat at 95% of the firms MIT NANDA studied, across 300-plus projects
Share of firms running generative AI pilots, by measured profit impact, MIT NANDA, July 2025, percent
The full circle is every firm in MIT NANDA's study running generative AI pilots, and the magenta arc is the 95% whose pilots left profit flat.
Source: MIT NANDA [1]. Chart by Ryan Elliott Dennis.
The numbers behind this chart
| Item | Value | Note |
|---|---|---|
| Pilots left profit flat | 95% | From 300-plus disclosed projects, 52 interviews and a survey of 153 leaders |
| Whole | 100% |
MIT Draws the Line Between Pilot and Profit
Raskar's report, "The GenAI Divide: State of AI in Business 2025," carries three co-authors: Aditya Challapally, Chris Pease and Pradyumna Chari1. Its 95% figure rests on more than one survey question. Analysts read more than 300 disclosed AI projects. They ran 52 interviews and polled 153 senior leaders at four industry conferences. Then they checked what people claimed against what the filings showed. Companies had already put $30 billion to $40 billion into generative AI by then. Five percent of integrated pilots had reached production or produced measurable value. Their central claim cuts against the industry's favorite excuse for slow returns. They blame how firms deploy the technology, ahead of model quality or regulation.
That difference matters for anyone reading enterprise-AI headlines closely. A gap caused by model quality would close on its own as labs ship better systems. Blame the deployment choices instead, and the cure becomes organizational change, which a model release rarely delivers. Raskar's team handed buyers a harder problem than waiting for the next model. Process, budget discipline and change management are the three things consultants charge the most to fix, which means the firms best placed to diagnose the 95% are the same firms that bill for the cure.
Gartner Prices the Coming Cancellations
The firm published its own forecast less than a year later, pointing forward rather than back. More than 40% of agentic AI projects running today will be canceled by the end of 2027, Gartner predicted in June 20252. It named rising costs, murky business value and weak risk controls. Verma's read matched MIT's almost exactly. "Most agentic AI projects right now are early stage experiments or proof of concepts that are mostly driven by hype," she said, adding that many use cases sold as agentic skip real agentic work. A January 2025 Gartner poll of 3,412 webinar guests caught the market mid-decision. 19% had made significant agentic-AI investments, 42% conservative ones, 8% had spent zero, and 31% stayed undecided. The mood moved too3. Generative AI for buying entered what Gartner calls the "trough of disillusionment" by July 2025, and The Economist had put that same label on the whole sector two months before, which makes the trough a shared finding rather than one firm's opinion4.
McKinsey Measures Conviction Against Cash
McKinsey's August 2026 survey shows MIT's divide outlasting one bad year. 37% of respondents credited AI with some EBIT impact, the same share as 2025, after another twelve months of buying, building and hiring against a technology their own finance teams could still find little trace of5. Only 6% qualified as "AI high performers,". That label needs 5% attributed EBIT impact plus significant reported value. Scaling numbers read better on their own. 44% reported AI running company-wide, up from 38%, and 88% reported regular use somewhere.
McKinsey named the tension itself. "Organizations' conviction in AI is growing faster than the immediate financial returns they can attribute to it," the firm wrote, a sentence that summarizes every report cited here5. 80% of respondents still reported better individual productivity, and 60% expected to raise AI spending anyway. Belief in the eventual payoff outlived the proof that it had arrived.
Deloitte Splits Productivity From Transformation
The firm fielded its 2026 State of AI in the Enterprise report across 3,235 senior leaders in 24 countries6. Results drew the same line between small proven gains and the bigger change firms say they want. 66% reported productivity gains, and 53% cited better insight and decisions. Real value at scale, even where it stayed small. Revenue told a harder story. 20% reported higher revenue from AI, against 74% who said they wanted it, a 54-point gap between ambition and result. 34% called their AI use deep change: new products, rebuilt processes, altered business models. Everyone else settled for efficiency or partial redesign. Momentum sits under those flat headline numbers. Deloitte expects the number of firms running more than 40% of their AI work in production to double within six months.
What the Chief AI Officer Actually Buys
Hiring patterns are the clearest answer to every statistic above. 76% of firms had a chief AI officer in 2026, up from 26% a year earlier7. IBM's Institute for Business Value surveyed 2,000 CEOs between February and April to get there. Nearly a tripling. Hiring rose in the same twelve months that three separate research houses published evidence of how little measurable return the spending had produced156.
Chief AI officers spread from 26% to 76% of firms in a year, while the share crediting AI with EBIT held at 37%
Firms with a chief AI officer, per IBM, and respondents crediting AI with some EBIT impact, per McKinsey, percent
Each pair of columns is one measure: the blue column is 2025 and the magenta column is 2026, so the left pair nearly triples and the right pair stays level.
Sources: McKinsey & Company (QuantumBlack) [5]; IBM Newsroom [7]. Chart by Ryan Elliott Dennis.
The numbers behind this chart
| Item | 2025 | 2026 | Note |
|---|---|---|---|
| Chief AI officer | 26% | 76% | IBM surveyed 2,000 CEOs between February and April 2026 |
| Credit AI for EBIT | 37% | 37% | McKinsey's August 2026 survey, the same share as 2025 |
Gary Cohn, IBM's vice chairman, named what changed for the people in those seats. "The CEO's role has always been to lead through disruption. What AI changes is the velocity and consequences of leadership."7 Read that second sentence against MIT's finding. Speed rose. Consequences rose. Measured profit impact stayed flat15.
A second hiring wave, for forward-deployed engineers, runs at the same firms commissioning these surveys. It reads as the fix MIT's diagnosis implies. If deployment explains the gap, then putting engineers inside business units beats buying more licenses. Whether that works is 2027's question, the one Gartner's forecast already prices as likely to fail on a large share of today's projects2.
Google Cloud's own survey data points the other way on the same question. Fielded in September 2025 across 3,466 senior leaders, it found 74% reporting a return on generative AI inside the first year, a gap from MIT's 95% figure wide enough to question how differently each report counted a pilot8. Fiona Tan, Wayfair's chief technology officer, backed that finding with a number of her own. "AI agents can be applied to so many use cases, the number of businesses adopting them should be 100%. I can quickly point to dollars saved," she told Google Cloud's researchers.
By the numbers
- Ninety-five percent of the firms MIT NANDA studied got zero measurable profit impact from AI pilots1.
- Thirty to forty billion dollars: the enterprise AI spending MIT measured against that 95% figure1.
- Over 40% of agentic AI projects will be canceled by the end of 2027, Gartner predicts2.
- McKinsey found 37% of respondents crediting AI with any EBIT impact in 2026, level with 20255.
- Six percent qualified as McKinsey "AI high performers,", a label needing 5% attributed EBIT impact5.
- Twenty against 74: Deloitte's percentage gap between firms reporting AI-driven revenue and firms wanting it6.
- Chief AI officers now sit at 76% of firms, up from 26% in 2025, per IBM7.
- One third of Deloitte's leaders, 34%, called their AI use genuine business transformation6.
What to watch
Deloitte's projected doubling gives the clearest near-term test6. If scaling keeps outrunning McKinsey's flat 37%, the 2027 surveys should show profit catching up to deployment5. Gartner's 2027 forecast sets a hard date for one question2. Does deployment discipline, the fix MIT prescribed, close the divide MIT found1? Chief AI officer hiring, already past three quarters of large firms, raises a third7. Does the role deliver the organizational change these reports say technology alone leaves half-finished?
Sources
- Aditya Challapally, Chris Pease, Ramesh Raskar, Pradyumna Chari, "The GenAI Divide: State of AI in Business 2025," MIT NANDA, July 2025, https://mlq.ai/media/quarterly_decks/v0.1_State_of_AI_in_Business_2025_Report.pdf.
- Gartner Newsroom, "Gartner Predicts Over 40% of Agentic AI Projects Will Be Canceled by End of 2027," Gartner, June 25, 2025, https://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.
- Gartner Newsroom, "Gartner Says Generative AI for Procurement Has Entered the Trough of Disillusionment," Gartner, July 30, 2025, https://www.gartner.com/en/newsroom/press-releases/2025-07-30-gartner-says-generative-ai-for-procurement-has-entered-the-trough-of-disillusionment.
- "Welcome to the AI trough of disillusionment," The Economist, May 21, 2025, https://www.economist.com/business/2025/05/21/welcome-to-the-ai-trough-of-disillusionment.
- McKinsey & Company, "The State of AI," McKinsey & Company (QuantumBlack), Aug. 25, 2026, https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai.
- Deloitte, "State of AI in the Enterprise," Deloitte, 2026, https://www.deloitte.com/us/en/what-we-do/capabilities/applied-artificial-intelligence/content/state-of-generative-ai-in-enterprise.html.
- IBM Institute for Business Value, "IBM Study: CEOs are Reshaping C-suite Roles for the AI Era," IBM Newsroom, May 4, 2026, https://newsroom.ibm.com/2026-05-04-ibm-study-ceos-are-reshaping-c-suite-roles-for-the-ai-era.
- Oliver Parker, "The ROI of AI: How agents are delivering for business," Google Cloud, Sept. 4, 2025, https://cloud.google.com/transform/roi-of-ai-how-agents-help-business.