Beyond the CLI: Agentic AI for Async Workloads and Non-Developers

Most of the conversation around agentic AI has focused on helping engineers write code faster. That's valuable, but it leaves a lot on the table. As coding accelerates, the bottlenecks elsewhere in the software delivery lifecycle become even more pronounced.

In this talk, we'll explore two complementary directions that extend the impact of agentic AI beyond interactive coding sessions. The first is expanding who AI agents serve: bringing agentic tooling to product managers, data scientists, designers, and other roles that are critical to software delivery but still underserved by today's tools. The second is async AI workloads: agents that run in the background to perform migrations, generate tests, triage bugs, and synthesize data — benefiting engineers and non-engineers alike, and requiring fundamentally different platform primitives than the synchronous copilot model.

We'll walk through what we're building at Airbnb, what we've learned, what's worked, and how to think about where these approaches actually move the needle.