How AI Agents Are Transforming Engineering Team Structures
Introduction: A New Era for Engineering Teams
On a recent Thursday evening in San Francisco, over a dozen events were hosted to connect AI startups with venture capitalists. Among them, Camp AI’s “Agents at Work” event, sponsored by Auth0, stood out. It featured companies that are actively reshaping their engineering workflows around AI agents. The event showcased vendors like Browserbase, Mastra, Fireworks AI, Drata, Mya, MindFort, and Corridor, all striving to make agentic AI secure and high-performing. Yet the most compelling narratives came from their own internal reorganizations—the successes and hurdles they encountered while restructuring teams for an agent-driven future.
Agentic AI Is Reshaping Team Structures
Smaller Teams, Bigger Scope
Paul Klein IV, founder and CEO of Browserbase, delivered a memorable line: “If AI is not doing your whole job it’s a skill issue at this point.” This reflects how rapidly AI is being woven into engineering practices. Abhi Aiyer, founder and CTO of Mastra, noted that organizations are now operating with dramatically smaller teams while tackling much larger projects. “You can have one person run a whole feature project because they have an army of one to infinity AI agents behind them,” he said. This shift means that a single engineer, supported by a swarm of agents, can accomplish what used to require an entire team.
The New Bottlenecks: Review and Approval
Several panelists pointed out that AI systems are generating code far faster than organizations can safely review and deploy. Aiyer explained that engineering teams are seeing a surge in pull requests, while the capacity to review them has become the critical bottleneck. Paul Klein stressed the need to throttle experimental AI output based on risk context. “If you are in the critical path and customer facing, no slop,” he emphasized. “If you are not critical path, not customer facing, slop away.” This risk-based approach allows teams to benefit from rapid generation without compromising production quality.
Trust and Ownership: Persistent Challenges
Accountability in Autonomous Systems
Observability and accountability were recurring themes. Rob Ferguson, VP of technology and strategy at Fireworks AI, argued that ownership cannot vanish simply because AI generated the output. “It doesn’t matter if you typed it or prompted it, you own it,” he said. Bhavin Shah, VP of AI product at Drata, highlighted the growing need for auditability in enterprise AI. “The agent is constantly telling the user, here is the action I’m taking, here is what I’ve done,” he noted. This transparency is essential for building trust in autonomous systems.
Securing the Agentic Workflow
Authentication and Authorization for Agents
Auth0’s demonstrations focused on authentication, authorization, and runtime controls for AI agents interacting with APIs and Model Context Protocol (MCP) servers. The company recently released its MCP authentication product to general availability, designed to secure agent interactions. Monica Bajaj, SVP of engineering at Okta, stressed minimizing risk exposure as agents operate autonomously across enterprise systems. “How do we make sure that those tokens are not long-lived tokens?” she asked, underscoring the need for fine-grained, short-lived credentials to prevent unauthorized actions.
Conclusion: Embracing the Agentic Transformation
The event made clear that reorganizing engineering around AI agents is no longer a future possibility—it is happening now. Teams are shrinking in size but expanding in capability, while bottlenecks shift from code generation to review and security. Trust and ownership remain frontier issues, demanding robust observability and accountability. As vendors like Auth0, Browserbase, and Mastra continue to refine their tools, the path forward involves balancing speed with safety. Engineering leaders who can navigate these changes will unlock unprecedented productivity, but only by addressing the new challenges that come with agentic AI.