AI agents in product teams — Centercode talk

What the video covers

Luke Freiler, CEO and CPO of Centercode, presents at the Orange County Product Managers meetup on how AI agents are entering product teams not as replacements for people but as a new category of coworker. The talk combines Centercode’s own experience adopting AI agents with Freiler’s observations from conversations with executives at large software companies about how they are thinking about the future of product teams.

Who it’s for

Product managers and product leaders who are curious about AI agents but uncertain about how to integrate them into team workflows. The talk is deliberately non-technical — Freiler focuses on organizational and practical implications rather than implementation details, making it accessible to PMs without engineering backgrounds.

Key takeaways

1. Agents are coworkers, not tools. Freiler draws a distinction between using AI as a tool (asking questions, generating text) and treating AI agents as team members with assigned responsibilities. At Centercode, the team names their agents, gives them cartoon avatars, and conducts reviews of their performance — practices that may sound whimsical but serve a practical purpose: they force the team to think about what each agent is responsible for and how to measure its contribution.

2. Deep research agents change what PMs can know. The ability to task an agent with continuous research — monitoring competitors, analyzing customer feedback, tracking market signals — means PMs can maintain awareness across a much broader information field than they could individually. Freiler describes this as “perpetual learning” and argues it creates a compounding knowledge advantage for teams that adopt it.

3. Digital twins let you test decisions before committing. Freiler describes building agent-based simulations of customer segments that allow PMs to test messaging, pricing, and feature positioning before running real experiments. These are not perfect predictors, but they compress the hypothesis-generation cycle and surface considerations the PM might have missed.

4. Product managers must become AI managers. As agents take on more team responsibilities, the PM’s role expands to include managing non-human team members. This means defining agent scope, setting quality expectations, monitoring outputs, and deciding when a task requires human judgment versus agent execution. The management skill is different from traditional people management but draws on the same principles of clarity, accountability, and feedback.

5. The comfortable-versus-new framework simplifies adoption. Freiler offers a practical heuristic: if you are comfortable with a task, enjoy it, and are good at it, keep doing it yourself and use AI as a sounding board. For tasks you dislike, struggle with, or find repetitive, delegate them to agents. This reduces the anxiety of AI adoption by framing it as augmentation of weak areas rather than replacement of strengths.

Worth watching if…

You manage a product team and want practical mental models for introducing AI agents as team members rather than as isolated tools, including how to evaluate what agents should and should not be responsible for.