AI coding agents can create pull requests faster than ever and the real bottleneck is no longer writing code. But we don't want to ship slop as fast as the agent can produce it.

At Wolt, a DoorDash company, we have been exploring how to move quality checks closer to the agent workflow, so AI does not just produce more code, but helps raise the bar for what reaches production. This talk shares practical patterns for closing the loop between agents, CI, code review, production monitoring, and team knowledge.

The talk will include concrete examples from Wolt-style engineering workflows. In particular, we will look at how AI-assisted development changes quality practices before, during, and after shipping: using agents for full-stack and cross-repository changes; improving PR reviews with project-specific context; and feeding production errors, logs, analytics, and rollout data back into future fixes.