I keep seeing conflicting headlines about Generative AI. One day it’s a revolution. The next it’s a failure. This paradox hit home recently with two major studies.
A new report from Wharton says 75% of enterprise leaders see a positive return on investment (ROI) from Gen AI. But I clearly remember an MIT study from earlier this year. Its headline claim was that 95% of AI projects deliver no measurable ROI.
So what’s going on? Are they both right? I decided to dig into the reports to understand the difference.
The Wharton Study: A Wave of Optimism
Wharton’s new report is called “Accountable Acceleration: Gen AI Fast-Tracks Into the Enterprise.” They surveyed around 800 senior leaders at large U.S. companies. The goal was to gauge their perception of Gen AI’s impact.
The findings are incredibly positive.
- Positive ROI is common: 75% of leaders report a positive ROI from their Gen AI investments.
- Budgets are growing: 88% expect their Gen AI budgets to increase next year.
- Usage is routine: 82% of leaders use Gen AI at least weekly.
But I noticed a small gap in the numbers. 75% of leaders report positive ROI, but only 72% formally measure it. This tells me some of this is based on a feeling. It’s informed optimism and perceived gains in productivity, not necessarily a hard line-item on the P&L. There’s nothing wrong with that, but it’s an important distinction.
The MIT Study: A Hard Look at the Numbers
The optimism from Wharton is a sharp contrast to the MIT Media Lab study. Its conclusion was blunt. 95% of the Gen AI projects they examined delivered no measurable impact on a company’s profit and loss (P&L).
Their methodology was different. Instead of surveying leaders about their feelings, MIT looked at individual projects. They wanted to see if a specific AI deployment actually changed the company’s income statement.
They found that most projects failed to move the needle for a few key reasons:
- Poor integration into existing workflows.
- The system didn’t learn or adapt to context.
- Firms aimed for “moonshots” instead of focused, narrow projects.
The projects that did work targeted specific back-office automations. Think accounting or finance operations.
So Who’s Right? It’s Apples and Oranges.
The truth is both studies can be right. They are measuring two completely different things. It’s a classic case of apples and oranges.
Here is the breakdown.
Wharton’s Study:
- What was measured? A leader’s overall perception of AI performance across their organization.
- How was ROI defined? Broadly. It includes productivity, efficiency, and a general sense of positive return.
- Who was asked? Senior leaders at large, mature companies already investing at scale.
MIT’s Study:
- What was measured? The financial performance of individual AI projects and pilots.
- How was ROI defined? Strictly. Did the project have a measurable P&L impact?
- Who was asked? A mix of deployments including public experiments where failure is more common.
Wharton is asking leaders “How do you feel about your AI investment overall?” MIT is asking “Did this specific project make or save us hard dollars?”
These are two very different questions.
What This Means For You
So how can you use this information? I think it comes down to being honest about your own AI projects. You need to ask a few simple questions.
- Classification Check: Which camp are you in right now? Are your AI wins based on a Wharton-style “perceived benefit” or an MIT-style “measurable P&L impact”?
- Evidence Test: If your board or an auditor asked you to prove your Gen AI ROI, what evidence could you show them? Would it be a feeling or a number?
- Action Question: If you’re in the Wharton camp, what would it take to move to the MIT camp? Can you track metrics better? Do you need to scale a project to see a real financial impact?
The goal is to move from a perceived benefit to a measurable one.
Key Takeaways
- Conflicting reports are common. A new Wharton study found 75% of leaders see positive Gen AI ROI, while an earlier MIT study found 95% of projects have no P&L impact.
- Methodology is everything. Wharton surveyed leaders on their overall perception. MIT analyzed the hard financial impact of individual projects.
- Both studies are useful. Wharton shows that leadership is optimistic and feels AI is working. MIT shows that turning that feeling into measurable financial gain is still very difficult.
- Be honest about your own ROI. Understand whether your AI successes are based on perceived productivity gains or on clear, measurable financial results. The goal is to bridge that gap.
Want to earn CPE for this topic?
- Compare Options: See how we stack up against others in our 2025 Flexible CPE Guide
- Understand the Format: Read how Nano-Learning works for CPAs.
- Check Your State: Ensure you are compliant with our State Requirements Guide.
Further Reading
- 95% of generative AI implementations in enterprise ‘have no measurable impact on P&L’, says MIT…
- The GenAI Divide: State of AI in Business 2025
- 95% companies have seen zero return on their AI investments and the reason is …, says study
- Annual MIT Sloan Management Review …
- Expanding Ai Impact With Organizational Learning Oct 2020
- Artificial Intelligence and Business Strategy
- Learning to Manage Uncertainty, With AI
Response
[…] Is Gen AI Actually Delivering ROI? Wharton and MIT Disagree. […]