Episode 570: The Enterprisification of Agents

In Episode 570 of *Software Defined Talk*, hosts Brandon, Matt Ray, and Kote dive into the growing trend of 'enterprisification' in AI agents, critiquing Google's recent keynote demo of agent-based systems for planning a marathon. They argue that while the demo showcased familiar enterprise patterns—agent registries, governance, observability, and identity—it ultimately amounted to little more than 'integration agents': lightweight LLM wrappers around existing microservices, not truly autonomous or transformative AI. The hosts contrast this with the more exciting vision of AI as virtual workers—autonomous, job-role-specific agents capable of end-to-end business tasks—suggesting that Google’s approach is overly cautious and focused on infrastructure over innovation. They also explore broader themes: the inevitability of enterprise software absorbing new technologies into standardized, controlled frameworks; the gap between flashy demos and real-world utility; and the need for deeper, more rigorous critique of AI’s economic viability, scalability, and long-term impact. The episode closes with a bonus after-show discussion on the necessity of using AI to be a credible critic, and the importance of moving beyond emotional reactions to ask hard questions about profitability, real-world adoption, and design intent.