How AI is pushing designers and engineers into one workflow

This article summarizes an episode of Aakash Gupta’s video series featuring Ed Bayes, design lead at OpenAI, and Gui Seiz, director of product design at Figma.

Image credit: Timmy Loen

When AI speeds up engineering, the traditional design process breaks down. Ed Bayes, design lead on Codex at OpenAI, and Gui Seiz, director of product design for AI at Figma, warn that leaders must completely rethink how their teams work.

Only adding AI to old workflows will not fix delays. To survive this disruption, product teams must bridge the gap between creating static layouts and building working software.

Changing the design process to keep learning

Waiting until the end of a project to write code wastes time. Since AI makes building a working prototype as cheap as drawing a flat layout, leaders must require teams to test ideas with live code right away. This means skipping slow approval steps.

This shift destroys the old step-by-step design process. In the past, a lack of time and money forced teams to slowly move from simple sketches to detailed models. However, that limit is gone.

Prioritizing velocity over speed
However, writing software instantly creates a dangerous temptation to skip talking to users. Moving fast in the wrong direction wastes money.

Bayes says the new challenge is knowing when to slow down. Teams must figure out when to move as fast as engineers and when to pause for deep design work, “so that you can have velocity rather than speed.”

Turning design files into shared workspaces

Building prototypes faster creates a new risk: disconnected tools cause companies to lose track of why decisions were made. Leaders must connect design files directly to live code. Without this link, fast code updates will erase the original design plan, making early files completely useless.

To stop this disconnect, teams use AI to bridge the gap. In the past, turning a live webpage back into a design file required hours of manual copying. Now, these tools sync instantly.

Bayes notes how spacing and borders match exactly. You can simply copy a link from a design file, paste it into an AI tool, and say, “update my code with the change I made here.” This creates an endless loop between designing and building.

Preserving institutional context
This direct link also protects company knowledge. When people quit, the reasons for their choices often disappear into hidden code updates. Seiz points out the frustration of inheriting work where “decisions got made in engineering that didn’t even get recorded.”

New tools fix this by connecting directly to the company’s official style guide, rather than just pasting flat shapes. This ensures new workers always build with the right rules.

Managing work when automated tools fail

Managing responsibility when job titles blur

Keeping clear goals when skills overlap