UI/UX Case Study · Retail · AI Vision
Replacing the "Manual Marathon" with AI Vision technology. Scan an entire basket in seconds without unloading.
Self-checkout promises speed but delivers frustration. Customers scan items one by one, bags trigger weight sensors, age verification halts checkout mid-transaction. Staff hovers nearby, ready to intervene. The technology works fine, but the experience was designed around the machine, not the human.
Invert the design. Instead of customers adapting to the machine, build a machine that adapts to customers. One action: place your basket. Everything else (recognition, age verification, payment) happens in seconds.
A kiosk with overhead AI cameras that recognize your entire basket in a single scan, combined with computer vision for age-restricted items. No weight sensors. No bagging area. No error loops. Works in existing Sainsbury's footprints.
You place a small item in the bagging area. The weight sensor doesn't register it. Screen freezes. "Unexpected item detected." A staff member has to walk over and verify what's in the bag. This isn't a glitch. It's the design. It happens dozens of times per hour across every self-checkout in every store.
Customer wants wine. Checkout triggers manual age verification. Staff must stop what they're doing, walk over, check ID, approve the transaction. What should take 30 seconds takes 3 minutes. And they're now dependent on staff presence, not independence.
Self-checkout was supposed to free up staff. Instead, they're permanently stationed at the error clearing station. Customers wait for them. Error messages pile up. The system that was meant to improve throughput actually reduces it.
Put your basket on the platform as-is. No unloading needed.
The AI scans everything in seconds. You just confirm what it found.
If you have alcohol in your basket, the camera checks your age automatically. No ID needed, no staff involved.
One tap to pay. That's it.
All items scanned. Age check for the wine is done automatically, no staff needed.
This is a conceptual project. No real-world testing was conducted. Here's what I'd measure if this moved to a pilot programme.
Like what you see?
I'm always looking for interesting design challenges. If you're working on something complex, I'd love to hear about it.
