Britannia A-Eye
Britannia A-Eye
Britannia A-Eye
CX design for an AI shopping assistant helping visually impaired shoppers navigate supermarkets independently.
CX design for an AI shopping assistant helping visually impaired shoppers navigate supermarkets independently.
Project Overview
Project Overview
For a visually impaired person, grocery shopping is rarely simple. Aisles become obstacles. Shelves become guesswork. Reading an expiry date or telling two similar products apart usually means relying on a stranger. Britannia wanted to change that, not by adding a feature, but by rethinking the shopping experience for a customer the retail industry has largely designed around rather than for. The ask was ambitious: build an AI assistant that could guide someone through a real, unmodified store, end to end, hands-free, using voice and camera as the only interface. My role was to work with a strategist to map the full customer journey, every phase, every decision point, every place the experience could break.
For a visually impaired person, grocery shopping is rarely simple. Aisles become obstacles. Shelves become guesswork. Reading an expiry date or telling two similar products apart usually means relying on a stranger. Britannia wanted to change that, not by adding a feature, but by rethinking the shopping experience for a customer the retail industry has largely designed around rather than for. The ask was ambitious: build an AI assistant that could guide someone through a real, unmodified store, end to end, hands-free, using voice and camera as the only interface. My role was to work with a strategist to map the full customer journey, every phase, every decision point, every place the experience could break.
For a visually impaired person, grocery shopping is rarely simple. Aisles become obstacles. Shelves become guesswork. Reading an expiry date or telling two similar products apart usually means relying on a stranger. Britannia wanted to change that, not by adding a feature, but by rethinking the shopping experience for a customer the retail industry has largely designed around rather than for. The ask was ambitious: build an AI assistant that could guide someone through a real, unmodified store, end to end, hands-free, using voice and camera as the only interface. My role was to work with a strategist to map the full customer journey, every phase, every decision point, every place the experience could break.
Project Type
Project Type
CX & Strategy
CX & Strategy
CX & Strategy
Year
Year
2025
2025
2025
My Role
My Role
CX Strategy & Journey Mapping
CX Strategy & Journey Mapping
CX Strategy & Journey Mapping
Client
Client
Britannia
Britannia
Britannia
Getting to a Solution
Getting to a Solution
Getting to a Solution
A feature wouldn't solve this. Adding a screen reader or a help button to an existing shopping app would have made the experience slightly less hard, not fundamentally different. The team's early thinking settled on a harder but more honest goal: remove the need for sight entirely, rather than design around its absence. That meant the solution couldn't live on a screen. It had to live in voice, in real-time spatial awareness, and in a phone's camera doing the work a pair of eyes would normally do. Once that direction was set, the question stopped being "what feature do we add" and became "what does an entirely voice and camera-led shopping journey actually require, step by step, to work in a real, unmodified store." That question is what the journey mapping work that follows was built to answer.
A feature wouldn't solve this. Adding a screen reader or a help button to an existing shopping app would have made the experience slightly less hard, not fundamentally different. The team's early thinking settled on a harder but more honest goal: remove the need for sight entirely, rather than design around its absence. That meant the solution couldn't live on a screen. It had to live in voice, in real-time spatial awareness, and in a phone's camera doing the work a pair of eyes would normally do. Once that direction was set, the question stopped being "what feature do we add" and became "what does an entirely voice and camera-led shopping journey actually require, step by step, to work in a real, unmodified store." That question is what the journey mapping work that follows was built to answer.
A feature wouldn't solve this. Adding a screen reader or a help button to an existing shopping app would have made the experience slightly less hard, not fundamentally different. The team's early thinking settled on a harder but more honest goal: remove the need for sight entirely, rather than design around its absence. That meant the solution couldn't live on a screen. It had to live in voice, in real-time spatial awareness, and in a phone's camera doing the work a pair of eyes would normally do. Once that direction was set, the question stopped being "what feature do we add" and became "what does an entirely voice and camera-led shopping journey actually require, step by step, to work in a real, unmodified store." That question is what the journey mapping work that follows was built to answer.
Mapping the Journey
Mapping the Journey
Mapping the Journey
The work started by breaking the experience into distinct phases: store arrival, onboarding, aisle navigation, product scanning, checkout, and repeat use at home. Each phase was mapped across four layers: the physical touchpoints involved, the actions the user takes, the exact AI dialogue at each step, and design recommendations for how that phase should actually be built. Mapping it this thoroughly mattered because a voice-only experience has no margin for ambiguity. A screen-based product can let a user glance, scroll back, or visually confirm something went wrong. This one couldn’t. Every phase needed a clear path forward and a clear way to recover if something didn’t go as expected, and that had to be decided in the journey map itself, not figured out later in development.
The work started by breaking the experience into distinct phases: store arrival, onboarding, aisle navigation, product scanning, checkout, and repeat use at home. Each phase was mapped across four layers: the physical touchpoints involved, the actions the user takes, the exact AI dialogue at each step, and design recommendations for how that phase should actually be built. Mapping it this thoroughly mattered because a voice-only experience has no margin for ambiguity. A screen-based product can let a user glance, scroll back, or visually confirm something went wrong. This one couldn’t. Every phase needed a clear path forward and a clear way to recover if something didn’t go as expected, and that had to be decided in the journey map itself, not figured out later in development.
The work started by breaking the experience into distinct phases: store arrival, onboarding, aisle navigation, product scanning, checkout, and repeat use at home. Each phase was mapped across four layers: the physical touchpoints involved, the actions the user takes, the exact AI dialogue at each step, and design recommendations for how that phase should actually be built. Mapping it this thoroughly mattered because a voice-only experience has no margin for ambiguity. A screen-based product can let a user glance, scroll back, or visually confirm something went wrong. This one couldn’t. Every phase needed a clear path forward and a clear way to recover if something didn’t go as expected, and that had to be decided in the journey map itself, not figured out later in development.
Presenting Britannia A-Eye
Presenting Britannia A-Eye
Presenting Britannia A-Eye
Britannia A‑Eye is an AI-powered voice assistant built on Google Gemini’s multimodal architecture and Vertex AI, with real-time mapping specific to the store it’s deployed in. Britannia partnered with Mithra Jyoti, a Bengaluru-based NGO, to roll the pilot out to 4,000 visually impaired shoppers, with the assistant pre-configured on their devices so there’s no onboarding friction at the store entrance. It’s currently live at one location, MORE Supermarket in Bengaluru, India, where it maps 100% of Britannia’s product range at that store. This isn’t a downloadable app. Because it depends on custom in-store mapping and real-time cloud data tied to a specific retail layout, it only works inside the locations it’s been built for.
Britannia A‑Eye is an AI-powered voice assistant built on Google Gemini’s multimodal architecture and Vertex AI, with real-time mapping specific to the store it’s deployed in. Britannia partnered with Mithra Jyoti, a Bengaluru-based NGO, to roll the pilot out to 4,000 visually impaired shoppers, with the assistant pre-configured on their devices so there’s no onboarding friction at the store entrance. It’s currently live at one location, MORE Supermarket in Bengaluru, India, where it maps 100% of Britannia’s product range at that store. This isn’t a downloadable app. Because it depends on custom in-store mapping and real-time cloud data tied to a specific retail layout, it only works inside the locations it’s been built for.
Britannia A‑Eye is an AI-powered voice assistant built on Google Gemini’s multimodal architecture and Vertex AI, with real-time mapping specific to the store it’s deployed in. Britannia partnered with Mithra Jyoti, a Bengaluru-based NGO, to roll the pilot out to 4,000 visually impaired shoppers, with the assistant pre-configured on their devices so there’s no onboarding friction at the store entrance. It’s currently live at one location, MORE Supermarket in Bengaluru, India, where it maps 100% of Britannia’s product range at that store. This isn’t a downloadable app. Because it depends on custom in-store mapping and real-time cloud data tied to a specific retail layout, it only works inside the locations it’s been built for.
How It Actually Works
How It Actually Works
How It Actually Works
Hands-free activation
Hands-free activation
Hands-free activation
A pre-registered shopper opens the assistant with a voice command on entering the store.
A pre-registered shopper opens the assistant with a voice command on entering the store.
Continuous scanning, not point-and-shoot
Continuous scanning, not point-and-shoot
The shopper holds their phone camera up and walks normally. Gemini's multimodal architecture processes the environment continuously, there's no shutter button and no need to perfectly frame a barcode.
The shopper holds their phone camera up and walks normally. Gemini's multimodal architecture processes the environment continuously, there's no shutter button and no need to perfectly frame a barcode.
Audio guidance
Audio guidance
Through the shopper's device speakers, the assistant acts like a friend walking beside them: "You are approaching the biscuit aisle," or "Move your hand slightly to the right to pick up the Marie Gold pack." This was one of the more deliberate design choices in the journey, the tone had to feel like quiet, ongoing company rather than a system issuing commands.
Through the shopper's device speakers, the assistant acts like a friend walking beside them: "You are approaching the biscuit aisle," or "Move your hand slightly to the right to pick up the Marie Gold pack." This was one of the more deliberate design choices in the journey, the tone had to feel like quiet, ongoing company rather than a system issuing commands.
Verbalising product details on contact
Verbalising product details on contact
Once a product is in front of the camera, the assistant reads out the variant, price, and any allergen or nutrition flags immediately, "This pack contains 24g of sugar," or "Warning: contains wheat allergens," without the shopper needing to ask.
Once a product is in front of the camera, the assistant reads out the variant, price, and any allergen or nutrition flags immediately, "This pack contains 24g of sugar," or "Warning: contains wheat allergens," without the shopper needing to ask.
Designing For What Could Go Wrong
Designing For What Could Go Wrong
Designing For What Could Go Wrong
A large part of the journey mapping work was anticipating failure states, since a voice-only experience has no visual fallback. If the assistant misreads a product, mishears a request, or the shopper goes quiet, there has to be a clear and reassuring next step. Every phase of the map included recommendations for how the experience should recover: tactile floor paths for physical navigation, haptic feedback before the AI speaks so users aren’t caught off guard, an SOS option for when a shopper needs human help, and a tone calibrated to stay friendly and low-pressure even when something goes wrong. A 35 to 40 second timeout phase exists specifically because silence in a voice-only experience is ambiguous. The shopper could be thinking, distracted, or stuck, and the assistant needs a graceful way to handle all three.
A large part of the journey mapping work was anticipating failure states, since a voice-only experience has no visual fallback. If the assistant misreads a product, mishears a request, or the shopper goes quiet, there has to be a clear and reassuring next step. Every phase of the map included recommendations for how the experience should recover: tactile floor paths for physical navigation, haptic feedback before the AI speaks so users aren’t caught off guard, an SOS option for when a shopper needs human help, and a tone calibrated to stay friendly and low-pressure even when something goes wrong. A 35 to 40 second timeout phase exists specifically because silence in a voice-only experience is ambiguous. The shopper could be thinking, distracted, or stuck, and the assistant needs a graceful way to handle all three.
A large part of the journey mapping work was anticipating failure states, since a voice-only experience has no visual fallback. If the assistant misreads a product, mishears a request, or the shopper goes quiet, there has to be a clear and reassuring next step. Every phase of the map included recommendations for how the experience should recover: tactile floor paths for physical navigation, haptic feedback before the AI speaks so users aren’t caught off guard, an SOS option for when a shopper needs human help, and a tone calibrated to stay friendly and low-pressure even when something goes wrong. A 35 to 40 second timeout phase exists specifically because silence in a voice-only experience is ambiguous. The shopper could be thinking, distracted, or stuck, and the assistant needs a graceful way to handle all three.
Reflection
Reflection
Reflection
Designing for voice and physical movement together is a different discipline from designing for a screen. A user can’t glance at something to course-correct, so every instruction has to land correctly the first time. There’s no visual reference to fall back on if it doesn’t. The hardest part of this work was deciding what the assistant should do and say in every moment things didn’t go as planned. That’s usually where trust in a system like this is won or lost. It’s the part of the design process I’d recommend the team keep refining as the pilot runs longer and the team learns more about how shoppers actually use it day to day.
Designing for voice and physical movement together is a different discipline from designing for a screen. A user can’t glance at something to course-correct, so every instruction has to land correctly the first time. There’s no visual reference to fall back on if it doesn’t. The hardest part of this work was deciding what the assistant should do and say in every moment things didn’t go as planned. That’s usually where trust in a system like this is won or lost. It’s the part of the design process I’d recommend the team keep refining as the pilot runs longer and the team learns more about how shoppers actually use it day to day.
Designing for voice and physical movement together is a different discipline from designing for a screen. A user can’t glance at something to course-correct, so every instruction has to land correctly the first time. There’s no visual reference to fall back on if it doesn’t. The hardest part of this work was deciding what the assistant should do and say in every moment things didn’t go as planned. That’s usually where trust in a system like this is won or lost. It’s the part of the design process I’d recommend the team keep refining as the pilot runs longer and the team learns more about how shoppers actually use it day to day.
Rights owned by VML India, WPP.
Rights owned by VML India, WPP.
Rights owned by VML India, WPP.
