By Greg Woolf, AI RegRisk Think Tank
Four months ago, an MIT report made waves with the claim that 95% of AI projects fail — hitting a nerve with AI fatigue, hype backlash, and a craving for caution. The study went viral, shaping the year’s boardroom narrative: AI might be exciting, but it wasn’t making money.
The problem? That MIT study, while thought-provoking, was based on a small convenient sample of 52 interviews and 153 surveys gathered from local conference attendees. It wasn’t designed to represent the broader market, and the authors were careful to label their findings “directional, not representative.”
Click Here to Learn More About the AI Readiness
Program for Financial & Wealthtech Firms
Now, the Wharton GBK 2025 AI Adoption Report has quietly rewritten that story with a far more rigorous, longitudinal study of roughly 800 enterprise leaders across multiple industries. Where MIT saw failure, Wharton finds accountability, discipline, and yes … even positive Return On their Investment (ROI)!
From Pilots to Profits
Wharton frames the evolution of enterprise AI in three stages: Everyday AI, Proving Value, and The Human Capital Lever.
AI has moved from novelty to workflow, from curiosity to core. 82% of enterprise leaders now use GenAI weekly and nearly half use it daily, evidence that adoption has gone mainstream.
But the real breakthrough is in measurement. For the first time, a majority of companies surveyed can show a clear return on AI investments.
- 74% of firms report positive ROI from GenAI initiatives with mid-sized companies leading the way.
- 72% now formally measure ROI, led by finance and HR, using the same rigor applied to other capital investments.
- 88% expect to increase AI budgets next year, with most anticipating payback within two to three years.
This is what Wharton calls “accountable acceleration” is the transition from experimenting with AI to managing it as a strategic asset.
Why the Wharton Report Matters
In a year dominated by hype and skepticism, Wharton’s research stands out for its credibility and scale. Now in its third year, the GBK study surveys hundreds of leaders using validated sampling, longitudinal tracking, and cross-sector weighting formulated as a genuine academic benchmark.
The MIT “95% failure” report, by contrast, was a limited exploratory study of 52 interviews and 153 surveys designed to understand why pilots stall, not how many succeed. Its authors acknowledged selection bias and a narrow time frame.
Wharton’s data doesn’t dismiss MIT’s findings; it contextualizes them. The “failure rate” by MIT described the pilot stage of early adopters. Wharton shows what happens when those pilots mature into real deployments: ROI appears, budgets grow, and accountability follows.
Everyday Use Cases and the Case for Moonshots
Wharton’s data shows that the biggest ROI today comes from “everyday AI” including the unglamorous but essential work of summarizing meetings, generating reports, analyzing data, and automating customer support. These efficiency gains may not make headlines, but they make CFOs and investors very happy.
Still, there’s a risk in stopping there. The companies that will truly win are pursuing a dual strategy: short-term ROI through operational efficiencies and long-term gains via transformation through moonshots, using AI to automate what was once unimaginable or prohibitively expensive.
These moonshots will reshape business models, redefine products, and create entirely new categories. Firms focused only on incremental gains may find themselves disrupted not by big tech, but by smaller players who move faster and use AI to reinvent the rules.
Measuring Expected Value and Thinking Like an Investor
For organizations still early in their AI journey, measuring ROI is only the first step. The next is adopting a risk-adjusted, expected-value approach by treating AI investments like a portfolio, not a gamble.
Here’s the framework forward-looking firms are using:
- Define the use case and baseline: Select one workflow, capture current costs, throughput, and risk exposure.
- Model scenarios: Estimate conservative, base, and upside outcomes for cost savings, productivity, and risk reduction.
- Assign probabilities: Calculate expected value using weighted outcomes.
- Adjust for risk: Discount returns for technical, adoption, and integration risk.
- Rank and fund like a portfolio: Prioritize projects that deliver the highest risk-adjusted ROI.
This approach borrows from M&A and corporate development playbooks where diversification, staged investment, and milestone-based funding are standard practice. If you don’t have internal strategists who think this way, there’s no shortage of consultants ready to help you develop the muscle.
From Curiosity to Core
If last year was about exploration, 2025 is about proof. The conversation has shifted from “Does AI work?” to “How does our ROI compare?”
The companies that manage AI like an investment, balancing operational payoffs today with transformative bets for tomorrow, will define the next phase of the AI economy.
Everyone else will be playing catch-up.
Because the truth is now clear: AI isn’t failing. It’s paying off and the smart money knows it.
Author’s Note
While originating the ideas and structure in this article, I used AI to handle the writing and grammar-checking as well as to analyze and extract key insights from the 26-page MIT State of AI in Business report and the 90-page Wharton GBK AI Adoption Report. The technology helped surface the most relevant findings that support the central question here.
Greg Woolf is an accomplished innovator and AI strategist with over 20 years of experience in founding and leading AI and data analytics companies. Recognized for his visionary leadership, he has been honored as AI Global IT-CEO of the Year, received the FIMA FinTech Innovation Award, and was a winner of an FDIC Tech Sprint. Currently, he leads the AI Reg-Risk™ Think Tank, advising financial institutions, FinTech companies, and government regulators on leveraging AI within the financial services industry. https://airegrisk.com
