AI has a knowledge problem

In the first episode of WTCK? – What the Company Knows, Stephan Ballenweg and Jakub Zavrel explore a core challenge for enterprise AI: company knowledge is scattered across documents, systems, and people’s heads—and even capable AI models can’t use information they can’t access or reliably retrieve. They discuss why search, company-specific terminology, secure access, and expert evaluation matter as much as the AI model itself. Their practical message: companies don’t need perfect data or infrastructure to begin. Start with a focused goal, connect relevant knowledge, evaluate results with experts, and improve iteratively. Key topics Why corporate knowledge is often fragmented, outdated, or hard to find Why smarter models alone don’t solve access to proprietary company information The role of search, retrieval, permissions, and company-specific context Why experts are needed to judge whether AI answers are accurate and complete How benchmarks and representative questions can help evaluate results Why capability gains may be a more meaningful measure of ROI than time saved How to start pragmatically without waiting for a perfect data architecture Why enterprise AI is an ongoing effort – not a one-time technical project Key takeaways Enterprise AI depends on access to relevant knowledge, not just model capability. Expert judgment and evaluation help make AI results more trustworthy. Companies can begin with a focused goal and improve their systems over time. Capability gains may reveal value more clearly than time savings alone. Aligning stakeholders and information needs is part of the work – not an afterthought.Creators & Guests Stephan Ballenweg - Host Jakub Zavrel - Host