AI takes center stage at KMWorld 2026
Three keynotes on the final day of KMWorld emphasized how AI technologies affect KM.
Samuel Anani, Head of Strategic Account Management, North America, at Behavox, discussed building defensible AI in compliance, noting that in regulated industries, implementing AI can have serious consequences if done wrong. Suppose, he asked, "If a regulator, a court, or your board asked how your AI reached an answer, could you show it's working?" You need to know how the answer is generated.
He stressed the difference between keyword searching, which searches for words not meaning, and AI searching, which excels particularly with unstructured data. Financial compliance, Anani added, is one of the hardest real-world tests of AI on unstructured data and can have billions of dollars in fines if out of compliance.
What's critical is having a taxonomy foundation, being able to search by meaning across languages and channels, explainability and provenance, and keeping knowledge current. The "right" kind of AI to use differs by task. For some, deterministic is appropriate, while others can be fully agentic. Anani stressed that you need a human reviewer on every compliance decision.
Anani ended by saying if you can answer these three questions, you are probably in a good space: Can you trace it? Can you explain it? Who signed off?
Mathieu Houdet, VP of Americas, AODocs, talked about how to architect trust when applying AI agents. One issue is when the highest semantic score doesn't align with the most recent document. Identifying the right document means the right version, the right permissions, the right context (metadata), and no hallucinations.
Houdet cited a bank customer who had a homemade RAG. Before AODocs, it didn't work very well, with only 60% accuracy. At a manufacturing company, AODocs did a benchmark test of Claude on File Share versus Claude with AODocs. The percentage of correct answers went from 71.8% to 94.5%.
Houdet ended by saying, "With great power comes the need for a solid foundation."
Jan Štihec, Director, Data & AI, Shelf, invited KM'ers to become the heroes of their company's AI by empowering knowledge experts to ground, guide, and shape AI outcomes.
"This is not a keynote, this is a call for action," he declared. Organizations take one of two paths: They put knowledge experts at the center to shape what AI knows and ground AI in what the business knows or they hack around the complexities of AI, leaving KM'ers on the sidelines, without influence or agency. Guess which one has successful pilots versus those that fail. Yes, it's the first path and that doesn't happen without KM involvement.
Someone needs to own the problem of AI's lack of understanding: Own the complexity, build AI's foundation, and scale with automation. It's knowledge experts who know how to do this. Language models don't understand your business, people do. To tame complexity, Štihec explained, you need structure, context, and quality. AI is our new consumer and that brings new requirements. Humans can judge when something is "off," but AI has no clue. Experts shape the knowledge so that AI can deliver, because many of the tasks cannot be achieved solely by hand.
Take action today to shape the complexity of your business, he urged. Establish the foundation and make sure your organization goes down that first path.
KMWorld returned to the JW Marriott in Washington D.C. on November 17-19, 2026, with pre-conference workshops held on November 16th.
KMWorld 2026 is a part of a unique program of five co-located conferences, which also includes Enterprise Search & Discovery, Enterprise AI World, Taxonomy Boot Camp, and Text Analytics Forum.