In Episode 193 of the AIAW Podcast, we welcome back Stefan Wendin, now Deployment Strategist at QuTwo, for a conversation about one of the most critical challenges in building reliable autonomous AI systems: agent memory management. As AI agents evolve from simple assistants into persistent, decision-making systems, we’ll explore how they remember, update, forget, and reconcile information over time, and what happens when their memories contradict what they have previously learned. Drawing on Stefan’s work with knowledge graphs, specialist AI models, and sovereign computing, we’ll discuss the shift from MLOps to AgentOps, the limitations of vector-based memory, and the engineering challenges of building trustworthy, stateful agents for enterprise environments. We’ll also examine how organizations can maintain control over their institutional knowledge, manage uncertainty and conflicting evidence, and what persistent artificial memory could mean for human judgment, digital sovereignty, and the future of human-AI collaboration.
