The Report “State of AI in Business 2025” Exposes MIT’s Blind Spots in Business and Enterprise AI Adoption

The headlines were loud and far reaching:

“95% of organizations are getting zero return on their AI investments.”

“Billions are being spent, adoption is high, but transformation is rare.”

This is what MIT calls the GenAI Divide, and it caused the stock of some major AI companies to take a hit. Except...

The big surprise is that MIT was surprised by these findings.

What the report framed as unexpected “findings” are in fact well documented realities of enterprise AI adoption. And this lack of contextual understanding throughout the report leads to weak conclusions and exposes some fundamental blind spots in MIT’s business and enterprise AI adoption analysis.

At Caspius, enterprise AI adoption is our focus. Here are the key blind spots we see in MIT’s report, along with a few useful reminders.

MIT Blind Spot: High Adoption, Low Transformation

One of the major “findings” from the report was that:

• 80% of companies have piloted AI, and 40% report deployments.

• But only 5% of companies report measurable value from AI.

This is nothing new. Research throughout 2025 shows the same: high enterprise AI usage, low ROI (hey, we even wrote about it here, here, and here). So, at best, MIT has confirmed existing research. The gap is the result of predictable issues: lack of strategy, poor data readiness, weak pilot design, and one-size-fits-all rollouts. And the nuance MIT missed:

AI usage ≠ AI adoption

Tinkering with AI tools to rewrite emails does not equal transformation. True adoption reshapes workflows, business processes, and outcomes at scale.

MIT Blind Spot: The Enterprise Paradox

MIT found that enterprises run the most pilots but take the longest to scale them. Mid-market firms (typically $50M-$500M in revenue) move from pilot to implementation in ~90 days. Enterprises take 9 months or more.

Again, not news. Anyone who has worked inside an enterprise understands this. Enterprise organizations are less agile and decision-making is slow. Organizations are layered, geographically dispersed, and often siloed. Workflows and business processes are complex. Scaling new technology is never quick in that environment.

MIT Blind Spot: The Shadow AI Economy

MIT “uncovered” and was apparently perplexed by what they call a “shadow AI economy.”

Findings:

• While only 40% of companies pay for official LLM subscriptions, 90% of employees already use personal accounts like ChatGPT or Claude for work.

Again, no surprise here. Reports from late 2024 and early 2025 found that employees are often ahead of their company in using Generative AI tools for their work.

The mistake MIT makes is assuming this individual usage translates to ROI. It doesn’t. True ROI remains at the workgroup level where AI technologies, generative or agentic, are integrated into the team’s work processes with proper access to corporate data.

At Caspius, our adoption model explicitly bridges this divide: aligning leadership-driven priorities with the AI use cases already emerging in the workforce.

MIT Blind Spot: Buying Beats Building

MIT concluded that internal builds fail twice as often as external partnerships.

Not surprising. Most corporate IT teams lack the depth and expertise to build AI solutions internally. For now, buying beats building. Enterprises will need external expertise until they can develop AI maturity in-house.

Where MIT Gets It Right

Despite the blind spots, some of the findings are useful reminders:

• Sectors with Higher ROI: Only Technology and Media show meaningful structural disruption. That makes sense. Tech leads naturally, and media is content-heavy, which aligns perfectly with GenAI.

• Where Budgets Go: 50–70% of AI budgets flow into Sales & Marketing. Metrics for these functions are easy to measure: demo requests, click-throughs, campaign conversions. But real efficiency gains are hiding in underfunded back-office functions like finance, procurement, and operations, because their value is harder to quantify.

Closing Thoughts

MIT’s report may have overstated the novelty of its findings, but it reinforces some truths:

• Usage ≠ transformation

• Budgets ≠ ROI

• Pilots ≠ production

And rather than chasing flashy demos, front-office metrics, or DIY builds, companies should focus on high-ROI use cases in their back office. The real wins are in integrating AI into the back office, aligning leadership priorities with workforce adoption, and moving beyond pilots into scaled, ROI-driven implementation.

At Caspius, this is where we focus: nuanced, enterprise-specific adoption strategies grounded in research and designed for measurable outcomes.

Feel free to reach out if you’d like to discuss where your organization stands on the adoption curve.