Introduction
For the past two years, we've been selling a simple promise: AI brings efficiency, reduces costs, and speeds everything up. I believe in that promise—it's what we do at Flexa every day. But after dozens of projects with clients from very different sectors, I've learned something that almost no one puts on their sales slides: if you use AI only to execute the same process faster, the promised return doesn't justify the investment. This isn't pessimism. It's what the data is showing.
The number that should bother us.
MIT published the study "The GenAI Divide " in 2025. Even with billions invested, 95% of organizations did not see a measurable financial return from generative AI. (Virtualization Review)
The most important part isn't the number—it's the diagnosis. MIT is straightforward: the cause isn't technological, it 's organizational; they call it the "learning gap," the inability to integrate AI into workflows, structure, and culture. The model works in the demonstration and fails in operation because the company around it hasn't changed.
McKinsey arrives at the same conclusion: a deep redesign of workflows is the factor most correlated with an impact on EBIT—but only 21% of companies have done this, while almost 80% simply overlay AI onto the old process. It's the "doing it faster" that automates inefficiency. Libertify
The 70% rule
BCG summarizes this in a formula that has become a benchmark: 10-20-70. Only 10% of the effort should go to algorithms, 20% to technology and data, and 70% to people and processes; more than two-thirds of transformations fail due to deficiencies precisely in this layer of change management. Neodata
Notice the inversion: most companies spend 90% of their energy choosing a model and supplier, and almost none on organizational change. That's why, at Flexa, we treat every AI project as a change management project , not as a technical deliverable—we redesign the flow before automating it and measure adoption, not just deployment. But this only works with real sponsorship from leadership.
The Brazilian perspective
The good news: according to AWS, Brazil leads AI adoption in Latin America . The bad news follows closely behind: only 12% of companies use the technology at a transformative level, compared to 62% in basic use—the challenge lies not in adoption, but in the depth of implementation. About Amazon Brazil
The Panorama 2026 survey , conducted by Amcham and Humanizadas, polled 629 senior executives and was spot-on: the difficulty in executing strategy, cultural resistance, and lack of prepared leadership even outweigh limitations in access to technology. Similarly, 63% of CIOs acknowledge a lack of maturity on the subject. (Terra Docmanagement)
Put the pieces together: we have access to the technology and the willingness to adopt it — but we're stuck on leadership and culture.
What does this change in practice?
If you're an executive evaluating an AI investment, three questions are worth more than any model comparison:
- Are we redesigning the process or just speeding up the current one? If the answer is "accelerating," the ROI probably won't materialize.
- Who owns the change? Not the owner of the tool — the owner of the behavioral change. McKinsey shows that the biggest obstacle is not the employees, but the leadership. McKinsey & Company
- How much of the budget goes to people and processes? If it's significantly below 70% of the BCG vaccine level, the design is unbalanced.
Conclusion
AI is one of the greatest productivity drivers of our generation. But it doesn't transform a company on its own—it amplifies what already exists, including disorganization and inertia. Technology has become cheap and accessible. Cultural change remains expensive, slow, and, unfortunately, optional for too many leaders. The turnaround will not come from those with the best model, but from those who have the courage to change the way the company works.
Sources: MIT The GenAI Divide: State of AI in Business 2025 · McKinsey State of AI 2025 · BCG AI at Scale (10-20-70) · AWS / About Amazon Brasil · Amcham + Humanizada Panorama 2026 · IT Forum.
