Designing teams for an agentic world
Summary
The traditional software organization structure, built on specialist scarcity, is becoming obsolete in an "agentic world" driven by AI, necessitating a fundamental redesign rather than merely adopting new tools. This shift is framed by four interconnected questions. Economically, organizations should prioritize buying commodity AI capabilities due to escalating training costs and decreasing inference costs, reserving in-house building for truly differentiating needs. This redefines talent, valuing expert generalists who orchestrate AI agents over fast coders, leading to smaller, senior teams. Structurally, delivery teams may form "pods" of 3-5 senior engineers, but the overall organization should adopt an "hourglass" shape to maintain a junior pipeline. At scale, a shared platform becomes crucial for around ten pods to manage security, compliance, and AI spend. Governance must adapt to probabilistic AI systems, requiring continuous oversight, provable agent identities, risk-category testing, and proactive strategies to combat deskilling and preserve human judgment.
Key takeaway
For Directors of AI/ML or VPs of Engineering tasked with integrating AI agents, you must move beyond bolting tools onto existing structures. Prioritize buying commodity AI and building only for unique differentiation. Focus your talent strategy on developing expert generalists who can orchestrate agents, not just write code. Structure your teams as senior-led "pods" within an "hourglass" organization to ensure a vital junior talent pipeline. Implement continuous, probabilistic governance for agents, including identity and risk-based testing, to manage uncertainties and preserve human expertise.
Key insights
AI agents demand a fundamental organizational redesign, shifting from specialist scarcity to generalist orchestration and probabilistic governance.
Principles
- Buy commodity AI, build differentiating capabilities.
- Value AI orchestration over raw coding speed.
- Preserve junior talent pipelines for future expertise.
Method
The article outlines a framework for organizational redesign based on four questions: economics (buy vs. build), talent (generalists with agents), structure ("pods" in an "hourglass"), and governance (probabilistic systems).
In practice
- Implement shared platforms for ~10 AI agent pods.
- Assign provable identities to all AI agents.
- Test AI systems by risk category continuously.
Topics
- AI Agents
- Organizational Design
- Talent Management
- AI Governance
- Software Economics
- Platform Engineering
Best for: Executive, Director of AI/ML, VP of Engineering/Data, CTO
Related on AIssential
See Counsel's argued verdicts on the open AI decisions leaders are weighing →
Editorial summary, takeaway, and curation by AIssential. Original article published by Anup Jadhav.