Over 75% of board directors and C-level executives are advancing their organizations’ AI strategies – and most of them are not getting the AI value out of their significant investment.
Comprehensive research streams daily into the market, highlighting the challenges in securing a positive return from AI investments, but this has yet to curtail the boundless enthusiasm to keep investing in AI. Many leaders have advised that this trend cannot persist for much longer, but despite the clouds foreshadowing, we are at a dangerous precipice where we could enter a winter period where AI is no longer the dominant attractor for digital transformation.
This blog dives into a summary of leading research sources identifying the failure rates in AI investments, from corporate research, industry market research, and academic sources. It includes recent research on what is needed to shift the tide to drive more success in realizing positive returns from AI and scaling AI investments into purposeful and productive operating models.
A detailed summary is also included of the recent McKinsey research on 12 transformation themes to alter this negative trend. The article concludes with Lady Whistledown’s thoughtful provocation and heeded warnings that await those who dare to walk the risky AI journey, where failure is more often the reality than success. Throughout this article, there are hidden gems to help you get AI right.
Notation: Our new book, The AI Precipice: An Executive Guide to Balancing AI Innovation and Safety is now available to order, and it is rich with guidance to guide C-level executives to achieve successful AI outcomes. If you reach out to me on LinkedIn and request an author’s discount code, I am happy to provide a 50% discount. I wrote this book with Cathy Cobey, the EY Global Head of AI Assurance, and it’s a go-to executive guidebook to increase your odds of success.
What are leading researchers saying about AI Failure Rates?
RAND Corporation
-
More than 80% of AI projects fail, twice the rate of non-AI IT projects.
Gartner
-
Approximately 85% of AI projects fail to deliver intended business outcomes.
-
Many failures are tied to a lack of business value and poor data readiness
McKinsey & Company
-
An estimated ~80% failure rate for AI initiatives
Generative AI: Even Worse Outcomes
MIT Media Lab / MIT Sloan (2025 studies)
-
95% of generative AI pilots show no measurable ROI or fail to scale
-
Only ~5% produce a meaningful financial impact
Abandonment & “Pilot Trap” Statistics
S&P Global Market Intelligence
-
46% of AI proofs-of-concept are abandoned before production
-
42% of companies abandoned most AI initiatives (2025)
Gartner
-
60% of AI projects lacking AI-ready data will be abandoned
-
~30% of GenAI projects are abandoned after proof-of-concept
Enterprise Value Gap (Adoption ≠ Impact)
Academic synthesis (2026, arXiv study)
-
Global AI investment: $252B+, yet only ~6% of firms see significant earnings impact
MIT / Enterprise studies
-
The majority of deployments deliver “little to no measurable impact” on P&L
Additional Supporting Industry Data
-
80%+ failure rate consistently cited across industry analyses
-
Up to 90% of AI projects may fail or be abandoned (industry forecasts)
-
Nearly half of AI initiatives at large firms are abandoned
Consolidated View
Across RAND Corporation, Gartner, McKinsey & Company, and MIT, the data converges:
-
80–85% of AI projects fail (baseline consensus)
-
Up to 95% fail in generative AI or pilot stages
-
30–60%+ are abandoned before scaling
-
Only ~5–10% deliver meaningful enterprise value
The Twelve McKinsey Themes Abbreviated.
McKinsey research (April, 2026) identified twelve themes that separate AI winners from their peers. These are summarized below, followed by a more detailed summary of perspectives.
𝟏. 𝐀𝐈-𝐂𝐚𝐩𝐚𝐛𝐥𝐞 𝐋𝐞𝐚𝐝𝐞𝐫𝐬. Business leaders, not IT, must own the AI agenda.
𝟐. 𝐀𝐠𝐞𝐧𝐭𝐢𝐜 𝐄𝐧𝐠𝐢𝐧𝐞𝐞𝐫𝐢𝐧𝐠. AI now runs complex autonomous workflows.
3. 𝐄𝐧𝐝𝐮𝐫𝐢𝐧𝐠 𝐂𝐚𝐩𝐚𝐛𝐢𝐥𝐢𝐭𝐢𝐞𝐬. Build lasting capabilities, not one-off solutions.
𝟒. 𝐎𝐫𝐠𝐚𝐧𝐢𝐳𝐚𝐭𝐢𝐨𝐧𝐚𝐥 𝐒𝐩𝐞𝐞𝐝. Embedding AI talent in the business shortens cycles.
𝟓. 𝐋𝐞𝐯𝐞𝐫𝐚𝐠𝐞 𝐏𝐨𝐢𝐧𝐭𝐬. Focus AI on your highest-impact areas.
6. 𝐕𝐚𝐥𝐮𝐞 𝐚𝐭 𝐒𝐜𝐚𝐥𝐞. Top AI companies see 20% EBITDA uplift by reinventing their core.
𝟕. 𝐏𝐞𝐨𝐩𝐥𝐞 𝐓𝐫𝐚𝐧𝐬𝐟𝐨𝐫𝐦𝐚𝐭𝐢𝐨𝐧. 70%+ of tech talent should be in-house and senior.
𝟖. 𝐏𝐥𝐚𝐭𝐟𝐨𝐫𝐦𝐬 𝐚𝐬 𝐒𝐭𝐫𝐚𝐭𝐞𝐠𝐲. Tech platforms drive speed and enable AI to scale.
𝟗. 𝐃𝐚𝐭𝐚 𝐚𝐬 𝐚 𝐁𝐮𝐬𝐢𝐧𝐞𝐬𝐬. Productize data, then enrich it continuously.
𝟏𝟎. 𝐀𝐝𝐨𝐩𝐭𝐢𝐨𝐧 & 𝐒𝐜𝐚𝐥𝐞. AI only creates value when widely adopted.
𝟏𝟏. 𝐃𝐢𝐠𝐢𝐭𝐚𝐥 𝐓𝐫𝐮𝐬𝐭. Agentic systems demand more rigorous testing and controls.
𝟏𝟐. 𝐂𝐨𝐧𝐭𝐢𝐧𝐮𝐨𝐮𝐬 (𝐑𝐞)𝐥𝐞𝐚𝐫𝐧𝐢𝐧𝐠. Skills expire fast. Those who relearn fastest win.
The Twelve Themes Explored
1. Enduring Capabilities over One-Off Solutions – Technology alone does not create advantage; the ability to harness it does. Successful companies, or “Rewired” companies, build enduring capabilities that allow them to outperform peers, rather than just delivering isolated AI tools.
2. Economic Leverage Points are Focal Points – Instead of chasing hundreds of use cases, successful companies focus on the few, core “economic leverage points” in their business models where AI can deliver massive, game-changing impact, such as supply chain, pricing, or R&D.
3. Value Must Move the Business – AI initiatives should be aimed at driving significant, measurable business impact. Leading AI adopters are achieving, on average, a 20 percent EBITDA uplift by concentrating on one to three key business domains.
4. Senior Leaders Must Drive the Tech Agenda – IT teams cannot drive transformation alone. Senior business leaders (one to three levels below the CEO) must be in the driver’s seat, owning the tech agenda and combining deep domain expertise with AI know-how.
5. AI Transformation is People Transformation – Successful companies increase their talent density by focusing on the “30-70 shifts”: over 70 percent of talent should be in-house, more than 70 percent should be “doer” engineers, and over 70 percent should operate at expert skill levels.
6. Speed is the Defining Advantage – The ability to quickly reallocate resources to high-value opportunities and reduce “latency”—the time from insight to action—is critical. Speed requires embedding AI talent directly into business units, not isolating them.
7. Treat Tech Platforms as Strategic Assets – Platforms should be managed as products with dedicated roadmaps and teams, rather than just backend IT infrastructure. These platforms accelerate execution speed and scale AI securely.
8. Productize and Enrich Data – AI requires massive amounts of high-quality data. Successful companies make data easy to consume by “productizing” it—treating data as a discoverable, accessible asset—and continuously enriching it for better performance.
9. Design for Adoption and Build for Scale – AI tools often fail because upstream/downstream processes aren’t changed. To succeed, companies must design for adoption by changing workflows and using modular, scalable architecture to deploy across markets.
10. No Trust, No Right to Deploy – With the rise of agentic AI, digital trust—protecting consumer data and ensuring cybersecurity—is the gatekeeper to adoption. Rigorous testing and automated guardrails are necessary for public and regulatory trust.
11. Master Agentic Engineering – Leading companies are moving from generative AI to autonomous agentic workflows. They are building the capability to integrate unstructured data, automate guardrails, and create a repeatable “agentic playbook”.
12. Learn, Unlearn, and Relearn – The half-life of skills is shrinking. Organizations that learn and adapt the fastest win. Leaders must take their top teams on “learning journeys” to understand the strategic opportunities of AI, rather than delegating the learning. An old adage that has meaning to achieve AI value realization is to ensure your leaders have “SKIN IN THE GAME.”
The Hard Truth for CEOs and Boards
If you strip away the hype, the research and experiences in the market are telling us that:
-
AI is high-risk capital allocation
-
Most organizations are not equipped to succeed
-
Failure is predictable—not accidental
The real question is not: “Are we investing in AI?”
Rather, it is: “Why do we believe we will be in the 10–20% that actually succeed?”
A Final Provocation for Your Boardroom – Based on the evidence, ask this directly:
-
If 70–90% of AI initiatives fail, what specific capabilities, governance mechanisms, and leadership behaviors make us confident we are not statistically destined to fail as well?
Our SalesChoice Promise
-
The work we do at SalesChoice is foundational AI strategy, AI governance, and sustainable AI solutioning.
-
We have learned the hard knocks of designing commercially graded and award-winning AI products, and we have helped many global organizations achieve positive ROI outcomes.
-
What sets us apart is our deep systems thinking and change management expertise that envelop every AI stage, guiding our customers’ AI journey forward.
-
We develop aligned organizational metrics that reward and recognize the right human behavior and design change stewardship networks to support the transformation needs.
-
We lay out the facts required to achieve successful AI outcomes. It takes courage to listen to our feedback and a willingness to upgrade operational practices.
-
We act as river guides. With humility and perseverance, we can tackle together the AI challenges, but only through collaborative and authentic kindness.
-
We will only be successful if we learn together and uplift each other – when we encounter a roadblock, actions will be swiftly taken that advance the foundations wisely laid.
-
It won’t always be easy – but we will get you to the other side, and together we will celebrate in the community building we have achieved and share our stories so others can learn from the footsteps we have dared to take.
Lady Whistledown’s Words of Heeding
Dearest Esteemed Directors and Most Ambitious Architects of Enterprise,
It has come to this author’s attention that the newest darling of the corporate court—Artificial Intelligence—is not, as so breathlessly proclaimed, a guaranteed triumph, but rather a most capricious suitor. Indeed, while many have rushed to embrace it with unguarded enthusiasm, the evidence suggests that a great many shall be disappointed in the match.
One hears whispers—persistent and increasingly difficult to dismiss—that the majority of these AI endeavors falter before they ever reach maturity. Grand pilots are launched with flourish, only to be quietly retired. Fortunes are invested, yet measurable returns prove elusive. And so, dear readers, one must ask: is it ambition that drives these pursuits, or merely the fear of being left behind in fashionable company?
For what is most curious is not that failure occurs—that, in itself, is no scandal—but that so many enter the arena unprepared for the discipline required to succeed. They mistake experimentation for transformation, and novelty for value. They speak of algorithms, yet neglect governance; they celebrate speed, yet disregard purpose; they pursue efficiency, yet overlook trust.
And herein lies the true source of the difficulty.
Artificial Intelligence is not a tool to be acquired, but a capability to be earned—through rigor, through restraint, and through a most unfashionable commitment to accountability. It demands not only investment, but introspection. Not only innovation, but judgment. Not only leadership, but stewardship.
You, dear Directors, are not merely observers of this unfolding drama. You are its guardians of consequence.
-
Will you permit your organizations to chase illusion—measuring success by activity rather than impact?
-
Will you accept metrics that flatter rather than reveal?
-
Will you allow governance to exist in name, but not in force?
Or will you insist—quietly, persistently—that every claim be tested, every risk be surfaced, and every ambition be anchored in something more enduring than trend?
For let it be plainly stated: the odds of failure are indeed formidable. Yet they are not indiscriminate.
Those who falter tend to share familiar traits:
-
They pursue AI without clarity of purpose
-
They invest without aligning operating models
-
They delegate accountability without retaining oversight
But those rare few who succeed—ah, they are far less impulsive.
-
They ask inconvenient questions.
-
They measure what others ignore.
-
They resist the temptation to declare victory too soon.
Most importantly, they remember that trust, once lost, is far more difficult to rebuild than any system is to deploy. And so, dear readers, take heart—not in the promise of easy success, but in the opportunity for distinguished leadership.
For in an age where many shall fail, excellence shall not require perfection—it shall require discipline. In a market crowded with proclamations, distinction shall come not from boldness alone, but from credibility. And in a world captivated by speed, the true advantage may well belong to those who proceed with clarity, purpose, and resolve.
Should you choose this path, you may yet find that what appears to be a perilous endeavor becomes something far more rare: Not merely a technological transformation… but a defining legacy.
Yours most observantly, Lady Whistledown
#AI #AIstrategy #AIgovernance #AIFailureStatistics #AIFailureRates #AIStatistics #AISuccessRates #BCG #Forbes #Forrester #Gartner #Gordon #LadyWhistledown #McKinsey #MIT #NTT #RAND#SalesChoice #Stanford #TheAIPrecipice #TheEconomist #TheGuardian
Research Sources
AI Research Aids
-
Gemini
-
OpenAI
AI Books
Institutional Reports
-
RAND Corporation. (2024). The Root Causes of Failure for Artificial Intelligence Projects. Retrieved from https://www.rand.org/pubs/research_reports/RRA2680-1.html
-
Gartner. (2024–2026). Generative AI Project Failure Rates and Enterprise AI Adoption Trends. Gartner Research & Press Releases.
-
McKinsey & Company. (2023–2025). The State of AI: Global Survey Reports on AI Adoption and Value Creation.
-
MIT Sloan School of Management & MIT Media Lab. (2025). The Generative AI Divide: Enterprise Adoption and ROI Realities.
Market Intelligence Reports
-
S&P Global Market Intelligence. (2024–2025). Enterprise AI Adoption, Pilot Abandonment, and Scaling Challenges.
-
Forrester Research. (2024–2025). AI Pilot-to-Production Conversion Rates and Enterprise Readiness.
-
Boston Consulting Group. (2024). AI Transformation: From Experimentation to Value Realization.
-
NTT Data. (2024). Global GenAI Report: Adoption, Investment, and Failure Trends.
Academic & Analytical Research
-
arXiv. (2026). AI Readiness Is an Organizational Learning Problem: Evidence from Global Enterprise Adoption.
-
Stanford University. (2024–2025). AI Index Report: Global Trends in AI Adoption and Impact.
Industry Commentary
-
Forbes. (2025). Why 95% of AI Pilots Fail and What Leaders Should Do Instead.
-
McKinsey: The AI Transformation Manifesto (April, 2026) Source: https://www.mckinsey.com/capabilities/tech-and-ai/our-insights/the-ai-transformation-manifesto
-
MIT Technology Review. (2024–2025). The Reality Gap Between AI Hype and Business Value.
-
The Economist. (2024–2025). Artificial Intelligence: Productivity Promise vs. Execution Reality.
-
The Guardian. (2025). Enterprise AI Adoption and Workforce Impact Trends.
-
TechRadar Pro. (2025). Why More Than Half of AI Projects Could Fail.
