Patrick Debois presents a thought-provoking exploration of how generative AI may reshape organizational structures through the lens of Conway’s Law – the principle that organizations design systems mirroring their own communication structures. Delivered at the “AI for the Rest of Us” event, the talk deliberately avoids deep technology discussion in favor of examining the human and organizational implications of AI adoption that apply across any role or department.
The talk begins with the familiar pattern of how new technologies land in organizations: one team incubates, others replicate, and eventually the capability scales out – often through a platform team. Patrick notes this is playing out with AI, with turf wars emerging between data science, application engineering, and platform teams over who owns the AI initiative. He highlights the “unfix” model’s experience crew concept, where a dedicated group ensures consistent AI user experiences across products, as Amazon discovered when their fragmented AI efforts needed alignment.
A central theme is the “reverse Conway” effect: as AI tools turn engineers into reviewers and managers of generated output rather than primary creators, roles and team structures must adapt. Patrick presents Sanjay’s framework for understanding how AI impacts jobs – from complementary augmentation, through competitive advantage, to commoditization and eventual substitution – and encourages the audience to identify where they can find new green-box value rather than being displaced. He references Henrik Kniberg’s observation that AI’s broad knowledge base may enable smaller teams, since individuals augmented by AI can cover more ground without needing as many specialists.
The talk takes a speculative turn when examining AI agents as organizational participants. Drawing on the Stanford “generative agents” paper, Patrick shows how multi-agent simulations mimic human social behavior and how companies are experimenting with agent-based software development teams, digital twins of organizations, and even formal employee records for AI workers. He raises the question of how organizations will handle agent performance reviews, codes of conduct, and the risk of one rogue agent corrupting an entire multi-agent workflow – mirroring the dynamics of toxic behavior in human teams.
Patrick closes with a provocative concept: “software-delivered services,” where AI agents can generate tailored software on demand, potentially disrupting the SaaS model the way 3D food printers might disrupt restaurants. While acknowledging that much of this is speculative, he frames the presentation as a spectrum of possibilities that organizations should be aware of as they consider how AI will influence not just their products, but the very structure of how they work.
Watch on YouTube — available on the jedi4ever channel
This summary was generated using AI based on the auto-generated transcript.