MIT Explained: Why Only 5% of GenAI Projects Generate Millions of Dollars in Value?

CONVERSATIONAL AI

Generative AI is one of the most powerful areas of transformation that the business world is rapidly trying to adapt to. This technology, which has entered everyone’s agenda thanks to tools like ChatGPT, represents not only efficiency for organizations, but also a redefinition of how business is conducted.

MIT’s “State of AI in Business” report offers valuable insights into how this transformation is progressing for organizations. According to the report, the majority of GenAI projects (95%) have not yet delivered the expected returns. However, this picture does not point to a systemic problem, but rather a transformation process that is still being learned. The same report shows that the correctly structured 5% of GenAI projects generate millions of dollars in value. In other words, success is possible—and the examples are clear.

At CBOT, we run AI projects with over 100 organizations. Our first step is always the same: to clearly identify the area that truly requires transformation. What problem are we trying to solve? How much does this problem impact user experience, operational efficiency, or financial outcomes? And most importantly, what is the most suitable technology for this transformation?

In this article, we share the core approach that enables GenAI projects to deliver visible outcomes, and explain how we manage this process.

Where Does Real Impact Begin?

Enterprise GenAI projects often start with big promises. But the real key to success lies in the first step: asking the right question.

Not “What can we do with this technology?” but “What problem are we trying to solve?”

A data-validated need can be more decisive than a powerful technological infrastructure. At CBOT, we begin every project with this mindset. We first listen to the process, evaluate it together, and then design the solution architecture. Because the same technology can yield completely different results in different contexts.

One key statistic from the MIT report makes this point clear: while 95% of GenAI projects create limited impact, only 5% deliver high value. This 5% represents projects that are process-centric, integrated into workflows, capable of learning, and context-aware.

Technology is certainly important. But what truly makes the difference is how that technology is used.

From Pilot to Production: Building the Right Architecture

Many organizations begin their GenAI journey with pilot applications—which is absolutely the right step. However, technical tests alone are not enough to move from pilot to production. The strategic framework must be aligned with the organizational context.

According to the MIT report, only 20% of organizations manage to move GenAI solutions to the pilot stage, and just 5% transition these projects to production. This data highlights a critical requirement for implementation: establishing structures that integrate with business processes.

At CBOT, we clarify the following questions at the beginning of every project:

  • Which business process will this solution integrate into?

  • How does it impact user experience?

  • What are the data flows and contextual variables within the process?

When these questions can be answered clearly, transitions become faster and smoother. Because GenAI is no longer a one-time software investment—it is a living, learning part of the organization.

The Quiet Dynamics of Enterprise Use

One of the most striking findings of the MIT report is that individual use has surpassed organizational transformation. At nearly 90% of companies, employees are already using generative AI tools for personal purposes. However, this usage often develops independently of corporate projects.

This actually points to a major opportunity. Individual usage preferences reflect the real needs on the ground. At CBOT, we analyze this “shadow economy” and incorporate real usage habits into project design. Which department is using which tool, for what purpose, and how? The answers to these questions make corporate roadmaps much more realistic.

A situation we frequently encounter in our pre-project analyses is this: the issue a company initially sees as top priority may drop to fifth place after analysis. This clearly shows the importance of data-driven decision making and focusing on real needs.

Directing Investment to Points of Impact, Not Just Visibility

In GenAI projects, the majority of investment is directed toward marketing and customer experience. MIT data shows that these areas receive nearly 70% of the budget. However, high return on investment (ROI) often appears outside these areas.

“Real transformation begins deep within operations.”

GenAI-based automations in back-office functions like procurement, invoicing, legal, or supply chain management not only deliver high efficiency but also reduce dependence on external resources.

At CBOT, the solutions we develop in these areas enable organizations to transform in a sustainable way. For example, the systems we implement in critical but invisible processes such as contract analysis, compliance checks, and collections deliver measurable results in a short time.

GenAI Myths: Confronting the Realities

When it comes to generative AI, discussions are often shaped by assumptions rather than data. However, MIT’s “State of AI in Business” report debunks five major myths with concrete evidence. The real picture is very different from common belief.

To build sustainable enterprise AI strategies, organizations must first confront these myths:

1. “AI will take our jobs”

A common concern is that AI will reduce employment. But the data tells a different story. The clearest observed impact so far is a reduction in outsourcing. Not mass layoffs—but a drop in the need for external support. This shows that instead of eliminating the workforce, AI provides an opportunity to use internal capacity more effectively.

At CBOT, we believe this transition should not be about systems replacing people, but rather about intelligent structures working alongside humans.

2. “AI is transforming all industries”

Transformation stories are often told as if they apply to every sector. But the reality is more cautious. So far, structural-level transformation has been observed primarily in the technology and media sectors. In other sectors, AI is still largely limited to pilot projects and narrow use cases.

With this awareness, CBOT avoids offering “one-size-fits-all” solutions. Instead, we build structures tailored to each sector’s maturity level and transformation capacity.

3. “Enterprises are slow”

A common but misleading cliché. It’s often said that large companies resist innovation. But MIT data says otherwise. Many large enterprises have already begun testing generative AI. The real issue is not speed, but implementation. In other words, there are plenty of ideas, but few functioning systems.

At this point, we consistently observe in our projects that speed alone is not enough—correct implementation architectures must be in place.

4. “The barriers are technical, legal, or data-related”

It’s commonly thought that technical or legal constraints are the biggest hurdles for GenAI projects. But the MIT report shows otherwise: systems that cannot learn, understand context, or adapt to the organization’s workflow are the ones that fail.

That’s why at CBOT, we begin every project by understanding the organizational context before focusing on technical feasibility. Technology only starts generating value when it’s built on the right foundation.

5. “We need to build our own AI”

Some organizations believe developing AI entirely in-house will give them more control. However, the MIT report reveals this approach is not as effective as assumed. Internally developed solutions are twice as likely to fail compared to externally sourced applications.

This data confirms a reality we also frequently encounter: some organizations prefer to run GenAI projects solely with their internal IT teams. This is a reasonable and natural tendency—especially for technologically competent firms. But we believe there is often a subtle yet critical aspect that’s overlooked: organizational know-how.

We are not just a technology provider. We bring a repeatable expertise gained from working with hundreds of organizations across dozens of industries. From prioritization strategies and scenario design to user experience and implementation sequencing, we offer extensive knowledge that makes organizational AI projects smoother and more successful.

Of course, it’s understandable that some companies want to move forward using their own resources. But in this journey, including not only technology but also experience enables stronger steps and faster, lower-risk transformation.

Any AI strategy developed without confronting these myths is built on risky assumptions. Decisions not supported by real data often cause initially exciting projects to be shelved quickly.

Success in Enterprise AI Transformation Relies on Clear Concepts, Realistic Expectations, and Context-Aware Applications

The New Era: From Static Tools to Dynamic Systems

One of today’s most critical concepts is: Agentic AI. These systems don’t just generate responses—they learn processes, adapt to context, and build reflexes. It’s no longer about producing the right answer, but taking the right action at the right time.

At CBOT, we place this approach at the core of our projects. Thanks to Agentic AI structures, organizations can build customized reflexes tailored to their specific needs.

The next step in these structures is Agentic Web—a new era where multiple AI agents can collaborate, coordinate, and carry out enterprise tasks by distributing responsibilities. This is made possible through multi-agent protocols such as:

  • MCP (Multi-Agent Coordination Protocols)

  • A2A (Agent-to-Agent Communication)

  • NANDA (Negotiation and Delegation Architecture)

With these systems, different roles within the organization can operate in sync. This means not just automation, but true organizational harmony.

Conclusion: Every Transformation Begins with the Right Question

Generative AI doesn’t just offer more data or speed. Its real value lies in building systems that are shaped by the right questions, integrated with context, capable of learning through processes, and developing organization-specific reflexes.

At CBOT, we don’t see AI as a product—we see it as the transformation itself. That’s why we view our projects not as software implementations, but as jointly constructed organizational architectures.

Many organizations today still perceive GenAI as a “technological option.” But we know that it is, in fact, a strategic structuring decision.

And every transformation begins with the right question.