AI Agency: Team Topologies as the New Infrastructure

Bridging Human and AI Agency

Matthew Skelton's argument that organizational maturity, specifically through the lens of Team Topologies, is the lynchpin for successful AI adoption is highly compelling. The core insight that 'bounded agency' for AI mirrors the principles already established for human teams is a crucial one. This framework provides a much-needed organizational structure to govern the inherent risks of AI, particularly the tendency towards 'excessive agency' and data exposure, which is a significant concern in the current LLM landscape. The comparison to OWASP Top 10 for LLM Applications further solidifies the practical relevance of this approach. The emphasis on 'stewardship' over mere ownership is a valuable reframing for managing complex AI systems over their lifecycle.

The article effectively highlights how organizations that have embraced Team Topologies will likely have a head start in integrating AI agents. The proposed solution of an 'Innovation and Practices Enabling Team' to diffuse successful patterns and the 'friendly FOMO' approach used by JP Morgan are excellent examples of how to foster cultural shifts and knowledge sharing, which are often the biggest blockers in technology adoption. The analogy to DevOps transforming software delivery is apt, suggesting that Team Topologies can similarly revolutionize AI integration. This perspective is particularly relevant as businesses grapple with demonstrating ROI from AI investments, suggesting that the problem is less about the AI itself and more about the organizational scaffolding around it.

Key Points

• Many businesses struggle to realize tangible returns from AI investments due to organizational maturity issues, not technical ones.
• Team Topologies principles, particularly 'bounded agency,' provide the necessary infrastructure for governing delegated AI initiatives.
• Granting AI 'unbounded access' to data is a critical vulnerability (Excessive Agency), akin to giving humans unchecked access.
• Organizations already structured for bounded agency in humans will find AI agent integration smoother.
• 'Stewardship' is a more productive framing than ownership for managing AI systems and services.
• Innovation and Practices Enabling Teams are crucial for diffusing successful AI patterns and fostering organizational learning.
• Active knowledge diffusion and shared success ('friendly FOMO') are more effective than mandates for large-scale AI adoption and cultural shifts.
• The evolution of AI necessitates organizational agility, with Team Topologies offering a framework to adapt.

---

📖 Source: QCon London 2026: Team Topologies as the ‘Infrastructure for Agency’ with AI