The AI Bubble: Billions Invested, Little Real Impact

According to MIT (“The GenAI Divide — State of AI in Business 2025”), 95% of companies have not seen significant returns from their AI projects.

Over the past two years, Artificial Intelligence has shifted from being a research topic to becoming a daily agenda item in companies. Tools like ChatGPT and Copilot have rapidly gained popularity, boosting the individual productivity of millions of professionals around the world.

But when we look closely at the numbers, a surprising fact emerges: 95% of companies that invested in AI have not achieved measurable returns.

According to MIT’s “The GenAI Divide — State of AI in Business 2025” report, despite investments between US$30 and 40 billion, only 5% of AI projects generate significant value on the P&L. The vast majority remain stagnant, with no real impact on operations or financial results.

This reveals a truth many developers and technology leaders have already noticed:
There’s a corporate “AI bubble” being fueled by hype, urgency, and indiscriminate adoption—often without a clear understanding of the problem to be solved.

The Blind Race for AI

In recent months, we’ve seen a true race to “have AI” in every work environment. However, many of these initiatives begin without answering three fundamental questions:

• What real pain point is this technology solving?
• What concrete return—financial or in efficiency—is expected?
• Is AI really the best solution for this context?

At 10i9, we believe the true value of AI lies not in speed, but in strategy.

It’s about mapping processes, understanding challenges, and applying AI in a precise, sustainable, and integrated way.
It’s about building solutions that generate real ROI, optimize time, reduce errors, and increase operational efficiency.
 AI is not magic. It’s a powerful tool—when applied correctly.

The Numbers Behind the Paradox

Generic tools like ChatGPT and Copilot are indeed boosting individual productivity.
However, they’re not delivering direct impact on companies’ financial performance.

And when we look at corporate AI solutions (custom or vendor-provided), the picture becomes even clearer:

• 60% of companies have evaluated corporate AI systems;
• Only 20% reached the testing phase;
• And just 5% made it to real production.

In short: there are plenty of pilots, but little real transformation.

AI Market Disruption Index

To better understand this scenario, a research team developed an index assessing sectors based on:

• Volatility of market share among leaders;
• Growth of AI-native companies founded after 2020;
• New business models driven by AI;
• Shifts in user behavior;
• Executive role changes linked to AI.

Results:

• Only two sectors—Technology and Media—show real signs of structural disruption.
• Professional services are seeing efficiency gains but limited deep transformation.
• Healthcare, retail, and financial services remain in pilot stages.
• Energy and heavy industries have barely advanced.

As one industrial COO put it:

“The hype on LinkedIn says everything has changed, but in our operations, nothing fundamental has.”

The Corporate Paradox

Large companies lead in the number of pilots, but lag in scaling.
This happens due to several structural reasons:

• Investment bias: Budgets favor “visible,” revenue-linked areas over the internal workflows that actually generate ROI.
• Lack of real disruption: AI is used as an accessory, not as a strategic engine of change.
• Fragile implementation: Internal projects fail twice as often as well-structured external partnerships.

Meanwhile, employees have already taken another path:

• 90% use AI personally at work,
• while only 40% of companies have purchased corporate AI subscriptions.

This “shadow AI”—simple, flexible, and bureaucracy-free—often delivers more practical value than many official projects.

The Problem Isn’t AI — It’s How It’s Used

Many executives still treat AI as a “silver bullet”, believing it will revolutionize entire sectors on its own.
But the real game-changer lies in learning and integration capabilities:

• Tools poorly integrated into real workflows are quickly abandoned.
• Lack of memory and adaptation limits AI to basic tasks.
• Disconnect from internal processes breeds organizational resistance.

When asked whether they preferred AI or a junior colleague for different types of tasks, the results were clear:

AI already dominates simple tasks. But to truly transform operations, it must learn, adapt, and evolve.

Companies on the Right Side of the “GenAI Divide” Have One Thing in Common

Organizations that are truly moving forward share a crucial trait:
They build adaptive systems, integrated into workflows and able to learn over time.

When asked what they really want from AI vendors, their answers were blunt:

“I’d rather wait for my current partner to add AI than bet on a startup.”

“Most don’t understand our approval flows or data.”

“If it doesn’t integrate with Salesforce or our internal systems, no one will use it.”

“The first week is great. Then it starts repeating the same mistakes.”

“If it can’t adapt every quarter, we go back to spreadsheets.”

Conclusion: Move Beyond the Hype, Focus on Execution

The AI race is full of noise, hype, and PowerPoint pilots that go nowhere.
But the organizations that are truly seeing results share a common approach:

• They know where to apply AI;
• They deeply integrate it into their operations;
• And they treat AI as a living, adaptive, strategic system—not a passing trend.

At 10i9, we believe the competitive edge doesn’t lie in simply “having AI,”
but in using it intelligently, connected to the business and oriented toward real outcomes.