According to a recent MIT report, 95% of GenAI pilots yield zero return. The 5% that succeed have figured something out about digital transformation in the age of AI. It’s a matter of focus, and knowing where to look.
Generative AI is often described in extremes: transformative, disruptive, revolutionary. Enterprises have responded in kind, pouring an estimated $30-40 billion into GenAI initiatives over the past two years.
Yet, according to MIT’s recent State of AI in Business 2025 report, fewer than 5% of AI pilots deliver meaningful transformation. The rest stall out — 95% of GenAI pilots yield zero return! — creating what researchers now call the “GenAI Divide”: a widening gap between companies experimenting with AI and those actually reaping value from it.
Technology itself is rarely the challenge; transformation is.
For executives, it’s a familiar problem. Technology itself is rarely the challenge; transformation is. Companies have little trouble launching pilots or rolling out tools like ChatGPT or Copilot. In fact, while only 40% of companies say they purchased an official LLM subscription, workers from over 90% of the companies surveyed regularly use personal AI tools for work in what the report calls the “shadow AI economy.”
What proves harder is embedding those tools deeply enough into the business that they change outcomes — reducing costs, increasing efficiency, or driving growth.
The MIT research shows why.
Most projects remain shallow integrations. AI tools are bolted onto workflows rather than woven into them, and as a result they fail to retain the context of a conversation or move fluidly across systems.
Too often leaders compound the problem by measuring adoption in terms of usage rather than transformation.
Many of the tools themselves lack the ability to learn and adapt. Systems that can’t admit when they’re not sure or lack the flexibility to refine their answers quickly lose the trust of users. And when back-end processes remain fragmented — order history in one database, billing in another, support tickets scattered across yet another — even the most well-meaning and well-funded customer-facing experiments crumble. Too often leaders compound the problem by measuring adoption in terms of usage rather than transformation, celebrating log-ins instead of cost reductions or improvements in service.
The winner’s mindset
By contrast, the small minority of companies achieving outsized returns approach the challenge differently.
They build AI into the core of their existing workflows rather than treating it as a surface-level add-on. They select systems designed to learn and improve over time, and they focus first on the high-friction processes where automation delivers the clearest returns: knowledge management, support workflows, document handling.
They also seek out strategic partners who can provide tailored solutions and deep expertise, and they evaluate success in terms of outcomes — speed, accuracy, efficiency, satisfaction — rather than raw adoption metrics.
What separates the 5% from the rest is discipline.
There is nothing magical about this. What separates the 5% from the rest is discipline. Where most organizations chase novelty, the leaders pursue integration, governance, and measurable performance.
Here lies the challenge for executives. Generative AI’s promise is real, but its success depends less on the sophistication of the model than on how well it’s integrated. The organizations that will define the next decade of AI adoption will be those that treat AI not as a product to deploy but as a capability to engineer, grounded in clear goals, robust data integration, and accountable results.
Getting results
Companies like Inbenta, which focus on knowledge engineering, workflow integration, and continuous learning, illustrate what this looks like in practice.
By unifying fragmented data, automating back-office processes, and designing systems that adapt over time, our solutions deliver meaningful experiences that drive growth.
Rather than offering AI as a stand-alone tool, Inbenta designs ecosystems where conversational interfaces, self-service tools, and human support can coexist. The goal is continuity — so a customer moving from a chatbot to a live agent doesn’t start over each time, and so employees can trust the system to surface accurate information when it matters.
Inbenta’s new platform, rolling out this fall, integrates these AI solutions even more closely in a single cohesive experience and a feedback mechanism that takes Inbenta’s +90% self-service rate and 99% response accuracy and improves them with each use.
Bridging the GenAI Divide
The lesson of the GenAI Divide is straightforward but sobering. Adoption is easy; transformation is hard. While the former may generate catchy headlines, it’s the latter that builds a durable competitive advantage.
And in the years to come, the gap between the two will only widen as the 5% pull ahead and the 95% keep falling behind.
In Brief:
- MIT finds that 95% of GenAI pilots fail to deliver ROI, creating a widening “GenAI Divide.”
- Shallow integrations, brittle systems, and fragmented back-end processes undermine most pilots.
- Too many leaders measure success by usage metrics instead of business impact.
- The 5% of companies that succeed focus on integration, adaptability, and measurable outcomes.
- Inbenta illustrates this approach by unifying data, automating back-office processes, and designing learning-first systems.
Inbenta makes AI simple.
Empower your users with AI-driven solutions across channels and content mediums — including voice, video, live translation, and seamless agent escalation.
