On August 18, 2025, Fortune published a hard-hitting piece: according to the MIT report The GenAI Divide: State of AI in Business 2025, 95% of generative AI pilots in companies are failing to deliver real business impact (read the full article here).

The main reason: too many initiatives remain stuck in the lab, never reaching production or generating tangible returns. CFOs and executives are growing frustrated: millions are being invested in AI, but results remain elusive.

At Baobab AI, this finding doesn’t surprise us. For us, it’s not a new trend—it’s the very gap we’ve been closing for more than a decade. Since 2011, we’ve been turning theory into practice, and practice into measurable impact.

From theory to practice

The MIT report highlights the gap between innovation and outcomes. Many companies experiment with algorithms, but very few manage to bring them into real operations.

That diagnosis matches exactly what Baobab AI has been solving for 14 years: translating mathematical expertise from academia into business impact.

• Where others showcase prototypes, we deliver measurable ROI.
• Where others run pilots, we deploy production-ready projects.

Real-world example: Paris Metro optimization

One standout case is our work with the Paris Metro on journey management.

Baobab AI led the development of a digital twin and mathematical modeling of the system, enabling scenario simulation and resource optimization for one of Europe’s most complex transport networks.

In the words of Laure Yang, project lead and automation & Big Data specialist at Metrolab (RATP and ATSA Subsidiary):

This project is a clear example of what the MIT report calls for: moving from experimentation to real operations. And Baobab AI achieved this over five years ago.

Economic impact across energy and industry

Beyond transportation, our Decision AI solutions have transformed industries:

• Iberdrola: optimized electricity distribution, improving resource allocation.
• Repsol: enhanced operational planning under uncertainty, cutting costs and boosting margins.
• Air Liquide and Nippon Gases: reduced planning times by 25% and improved logistics resource utilization.

These cases illustrate the difference between a failed pilot and a production solution with direct bottom-line impact.

Decision AI vs. Generative AI: impact, not promises

While generative AI dominates headlines yet struggles to show up in financial statements, the Decision AI we apply at Baobab AI focuses on what CFOs and executives truly care about:

• Real cost savings
• Optimal, traceable decisions
• Measurable operational efficiency

Our DecisionOps platform makes it possible to deploy Decision AI projects quickly and with real business impact. It integrates data from multiple sources, manages advanced optimization algorithms, and enables “what if” scenario comparisons. All of this with full production deployment capability, in the cloud or on-premises, and complete decision traceability.

This isn’t a prototype. It’s a tool that helps our clients make better decisions every day.

Over a decade of getting it right

The MIT report and Fortune’s coverage confirm that most companies are still figuring out how to apply AI with real impact.

At Baobab AI, we can point to 14 years of doing it right, with transparency, ethics, and verifiable results across demanding sectors such as energy, transport, and industry.

Most AI projects fail to deliver impact, as the MIT confirms. But there are players who turn theory into real practice.

At Baobab AI, we don’t chase tech trends, we’ve always focused on Decision AI to enable the right decisions: measurable, traceable, and profitable.

While MIT and Fortune show that most companies are still stuck in failed pilots, Baobab AI has been delivering successful, tangible AI results for more than a decade.

Better decisions are possible. And making them rigorously is too.