Building a Generative AI Design Deviation Tool in a Day
Built an AI-powered tool to automate design deviation detection between Figma files and staging builds, unprompted and in a single day.
Akhilesh Joglekar
Associate Designer at Walmart Global Tech India
From their time as
Associate Designer
Walmart Global Tech India • 2023
Overview
Akhilesh was tinkering with AI agents inside Walmart's internal tools when a colleague tapped him on the shoulder and asked a simple question: could AI identify design deviations between a Figma file and a staging build?
The Story
Akhilesh was tinkering with AI agents inside Walmart's internal tools when a colleague tapped him on the shoulder and asked a simple question: could AI identify design deviations between a Figma file and a staging build?
He did not know the answer. He said he would find out.
He spent the next several hours working through the problem using Walmart's internal AI tools. He mapped out what the tool would need: access to the staging environment, Figma file tokens, and a structured output format. He designed the report schema himself, specifying that the tool should identify each deviation, explain why it existed, describe how to fix it, and grade the severity of the issue.
The tool worked. It compared design screens against staging builds side by side, produced a detailed page-by-page deviation report, and graded each issue by urgency. A process that had previously taken designers a full day or more for a complete flow was reduced to an automated report.
The tool started as a proof of concept. The reorg happened the next day, and Akhilesh was laid off before he could hand it off or take it further.