Amazon Go Autonomous Retail

I joined as a founding designer, which meant building from first principles without existing patterns or a design team in place. Everything from entry flows to back-of-house moderation tools was defined on this project.

  • Founding UX designer, end-to-end customer experience and end-to-end associate experience

  • Designed across: Amazon Go mobile app, in-store tablet and kiosk displays, bidirectional smart food scale, digital signage, entry and exit flows, planogram, and human-in-the-loop moderation tools

  • Partnered with engineering, computer vision, retail ops, grocery leads, industrial designers, store designers, packaging designers, and design technologists

My results:

  1. Just Walk Out tech scaled to 140+ third-party locations in US, Australia, and Canada.

  2. 18M+ items sold in Just Walk Out stores.

  3. Lumen Field, home of the Seattle Seahawks and Seattle Sounders, reported an 85% increase in transactions and a 112% increase in sales, per game.

  4. Market Express at ExCel London, which has 400 events and more than 4 million visitors a year, is now able to serve 300% more customers on their busiest days and has increased annual revenue by 56%.

Customers had never shopped in a store with no checkout. There was no mental model to design for and no prior interaction pattern to borrow from anywhere in retail.

The experience had to work across physical space, computer vision, and a mobile app three systems that had never been designed together at retail scale.

The Situation

The problems I was accountable for solving:

  1. Customers had no legible model of what the system was tracking or when. Live receipt accuracy lagged behind real-time item detection, creating anxiety during the shopping experience

  2. Associates needed to detect customer account issues and dietary preferences in the moment, without interrupting the experience

  3. Human-in-the-loop reviewers needed tooling to spot-check transactions and resolve disputes accurately and efficiently

  4. The interaction model had to scale across multiple store formats (small convenience, medium grocery, and large-format) without breaking the core trust framework

Set a digital design foundation instead of executing narrowly on individual features

Early pressure was to move fast on individual surfaces. Instead I established a shared foundation that could support parallel work-streams: customer app, associate tooling, back-of-house systems, without each team building independently and incoherently. This cost depth in any single area early on, but it set a higher quality bar across the full experience, scaffolded a shared vision, and created the structure needed to hire and onboard a as the team grew.

The Strategy

Make it easy for customers to dispute charges

Dispute tooling was initially scoped as a back-office function. I advocated for surfacing it prominently in the customer-facing receipt experience. The reasoning: customers who can easily flag an error are more likely to trust the system overall, and dispute data directly improves the accuracy of human-in-the-loop review. Friction in dispute flows erodes confidence in the technology.