Are Uber's Agentic Pods the ultimate AI-native team setup? I'm such a fan that I think a book needs to be written about this.

Uber’s CTO Praveen Neppalli Naga recently shared their concept of “Agentic Pods” and I think every company should be studying some version of this.

They handpicked 30 of their most AI-proficient engineers and paired each one with a domain expert from a business function.

The first step is not to immediately build an agent. It is to sit next to the expert, shadow the work, observe every step, document the workflow, ask questions, and build intuition.

Then you prioritize the best opportunities based on four things: scale, repetition, business impact, and data availability.

Is this happening at high volume? Is it repetitive? Does it save meaningful cost or create growth? And most importantly, is the data actually available? Because if the data is not available, the agent will not work.

Then the engineer and the domain expert build the working agent together, for that specific use case.

Then the final step is validation: test the agent with other people doing similar work and see if it generalizes. If it does, you can clone it, adjust the skill file, change the MCP server, tweak the prompt, and make it work for other teams or workflows.

I’m very pro companies learning to build their own agents instead of just buying generic agent products from vendors, because so much of this work is custom to how the company actually operates.

For the people inside the company, learning how to build and manage agents is incredibly valuable. You make yourself more irreplaceable, more unfirable, and infinitely more hireable when you stop relying on others to automate your work for you.

That is the real AI-native team setup.

Bora Celik

Founder, Gentic

—-

Reply to this email or DM me on LinkedIn to discuss your brand's agentic transformation via gentic.co