GenAI Failing? Your Employees Are the Missing Piece, Not Data Scientists
Many people think AI success comes down to having cutting-edge algorithms or a high-powered team of AI/ML engineers and data scientists — but that’s not the case.
Research from top universities and global consulting firms shows that up to 80% of the work in AI projects isn’t about model building at all. It’s about getting the data right — cleaning it, organizing it, and making sure the right knowledge is accessible.
Without that solid foundation, even the most powerful AI can’t deliver real business value. Studies from Harvard, MIT, Stanford, Oxford, and firms like McKinsey, Deloitte, and Gartner all point to the same conclusion: the real challenge in AI isn’t technology — it’s the clean organizational knowledge required to feed the AI.
The smartest way to approach getting your company's knowledge ready for AI is by equipping your employees with the right tools and foundational data science training—because no one understands the data better than the people already working with it. Employees understand the business context, know where the data lives, and can spot what matters and what doesn’t.
Tapping into the domain expertise already inside the organization will consistently outperform external consultants or new data scientists who lack firsthand experience and context with the knowledge & data. Instead of bringing in new hires, focus on helping current staff understand how to structure, clean, and interpret data, along with basic principles of how AI systems learn from it—which are all components in an embedded Knowledge Management program.
Data scientists or external consultants might know the technical algorithms and creating complex data-workflows, but they often lack the institutional knowledge to interpret patterns, identify quality gaps, or prioritize what’s useful. They're hired to create insightful algorithms and models, but actually end up spending 80% of their time trying to find, fix, understand, and clean the data (InfoWorld).
As I found working in many Fortune 500 enterprises - what starts off as a highly organized location for content, over the years, turns into a dumping ground for knowledge assets that are incomplete, duplicate, rough drafts, obsolete. We found the best solution was to train (or refresh) the employees who worked with that content daily on KM best practices.
That’s why training internal domain experts in the basics of data structuring, labeling, and contextualization is essential. They don’t need to become coders or data scientists— they need to understand how their knowledge connects to what AI systems can learn from. This is the heart of modern knowledge management: transforming human expertise into structured, machine-readable insight. Enterprises who understand this will win with AI.
