How to Get Your Org to Adopt Coding Agents (Without Shipping Garbage)
AI coding agents promise 10x. On complex, production work inside a real org, the honest number is 2-5x — and getting there requires a journey most teams aren't prepared for. At Figma, we ship AI products to millions of users, but internally our engineering org is spread across three stages of adoption. The honeymoon, where AI is magic. The crash, where AI writes bad code and your best engineers are stuck protecting the quality bar. And the real skill — 2-5x with disciplined development practices and proper investment. This talk covers why adoption is uneven, what the trust curve looks like from the inside, and what leaders can do about it: guide teams to align on plans before generating code, set honest expectations, invest in the fundamentals that make codebases agent-friendly, and create space for skeptics without judgment. You'll leave with a framework for driving adoption more organically without mandating it — and without shipping garbage.
About the AI-Native Enterprises Track
AI-Native Enterprises sessions at AI Engineer World's Fair 2026 in San Francisco.
When
Tuesday, June 30, 2026
3:20 PM - 3:40 PM·20m
Where
Leadership 1 · Room 3016
Capacity: 550 attendees
Sign in to add this talk to your schedule.
Speaker
Eyal Blum
Figma
Eyal Blum is a senior staff engineer at Figma working on Client Testing, Observability, and Performance. He works on developer infrastructure and drives AI-assisted development practices across the engineering org, sitting at the intersection of developer tooling and AI adoption — figuring out what it takes for coding agents to actually work in a large, established codebase. Before Figma, Eyal spent nearly 20 years building large-scale systems at companies like Meta, Dropbox, and Google.