E167 - Practical Limitations for Agentic Business Processes - Luka Crnkovic-Friis

Description

In Episode 167 of the AIAW Podcast, we’re joined by Luka Crnkovic-Friis, Head of AI/ML at King (Microsoft), for a sharp and timely conversation on the practical limitations of deploying agentic AI systems in real-world business processes. Drawing on his leadership in scaling AI at the company behind Candy Crush, Luka dives into the critical bottlenecks businesses face when integrating reasoning models and autonomous agents—from accuracy trade-offs and productivity pitfalls to the psychological challenges of human-AI collaboration. We explore the shift from System 1 to System 2 AI, the rising need for deeper contextual understanding, and how Reinforcement Learning could unlock the next frontier of enterprise automation. If you're navigating the complexity of LLMs, decision-making agents, and the future of scalable AI in business, this is a must-listen episode on what’s possible—and what’s still holding us back. 00:00 Intro: Agentic AI, Practical Limits & the Future of Work 00:08 Luka Crnkovic-Friis: AI Leadership at King and Microsoft 00:14 Real-World Challenges Deploying AI Agents in Enterprise Operations 00:43 System 1 vs System 2 AI: LLMs, Reasoning & Reinforcement Learning 01:26 Human Context in AI: UX Challenges and Cognitive Trust Gaps 01:40 Can Agentic AI Stay Accessible as Complexity Grows? 01:58 Frontier Model Architecture: What’s Next for Advanced AI Systems? 02:12 Fast Takeoff Scenarios: AI Ethics, Governance & Societal Risks 02:20 AI Divide and the 95% Use Case Gap: Insights from MIT Research