Crossing the GenAI Divide

What AI in Business Teaches Us About Knowledge Systems

The recently released State of AI in Business 2025 report has been making waves with its stark conclusion: despite billions of dollars invested, 95% of organizations are realizing little to no measurable impact from generative AI initiatives. The authors call this gap the GenAI Divide — a fault line between organizations that experiment with AI and those that actually achieve transformation.

At first glance, the report seems to be about AI adoption. But read deeper, and you see something more fundamental: it’s about knowledge. It’s about how organizations fail — and sometimes succeed — at embedding new forms of learning into their systems. And that lesson applies as much to knowledge management as it does to artificial intelligence.

---

High Adoption, Low Transformation

According to the study, most organizations have eagerly experimented with tools like ChatGPT and Copilot. Adoption rates are high. Pilots are everywhere. But when it comes to business transformation, the results are almost nonexistent. Only 5% of enterprise AI pilots cross the chasm into scaled, profitable systems.

This should sound familiar to anyone who’s worked on a knowledge management initiative. We’ve seen it before: companies buy expensive platforms, set up new repositories, and launch flashy portals — only to find that real work continues elsewhere. Adoption is not the same as transformation.

---

The Learning Gap

The report pinpoints the barrier: most AI systems don’t learn. They generate text, automate tasks, and provide temporary efficiency, but they don’t adapt to context, retain memory, or evolve with the organization. In other words, they don’t become part of the organization’s living knowledge system.

Knowledge management initiatives fail for the same reason. It’s not enough to store information or even make it searchable. Unless systems evolve with feedback, unless they adapt to workflows and become woven into practice, they remain unused.

I’ve often argued that knowledge emerges through conversation, not documentation. The same is true for AI: without feedback loops and contextual learning, these tools stay static. They remain information systems, not knowledge systems.

---

Shadow AI and the Human Workaround

One of the most striking findings of the study is the emergence of what the authors call the shadow AI economy. While official enterprise projects stall, employees are quietly using personal AI accounts to get real work done. Nearly every professional surveyed reported using AI tools independently, even if their company had not officially adopted them.

This mirrors what happens when knowledge management systems fail: people route around them. They create informal Slack channels, swap spreadsheets, or build their own wikis because the official systems don’t fit the way they work. Knowledge finds a way, whether leaders sanction it or not.

The lesson? Pay attention to these grassroots practices. They reveal where knowledge grows naturally — and where official systems are blocking it.

---

How to Cross the Divide

The organizations that do cross the GenAI Divide share a common approach: they demand systems that learn, adapt, and integrate deeply into workflows. Instead of shiny demos, they look for process-specific customization. Instead of general-purpose tools, they invest in narrow but high-value use cases.

The parallel in knowledge management is obvious. The systems that work aren’t the ones that aim to capture everything, everywhere. They’re the ones that embed into the flow of work, enable learning over time, and build trust through visible value.

---

From AI Systems to Knowledge Systems

What this study really tells us is that the future of AI — like the future of knowledge management — hinges not on adoption but on adaptation. Organizations that thrive will be those that:

• Create feedback loops where tools and people continuously learn from each other
• Prioritize integration into existing workflows over standalone platforms
• Respect informal practices and build on the shadow systems employees already use
• Focus on augmentation, not replacement, designing AI and knowledge systems to extend human capabilities, not sideline them

When we view AI as part of the broader challenge of knowledge systems, the GenAI Divide becomes less a story of failed technology and more a story of organizations struggling to learn. And that, ultimately, is the real lesson: transformation only happens when systems — human and machine alike — are capable of continuous learning.

---

What about your organization? Are you still piloting tools that don’t adapt, or are you building systems that actually learn with you?