Product
Three interfaces. One connected store.
OnSight is a proposed software-based service and intelligence layer: shopper access, staff response and retail insight.
01 — Shopper
Help when you want it. Autonomy when you don’t.
OnSight is designed to let shoppers search products, check sizes and colours, locate items and ask for help on their terms, through in-store screens or proposed QR/mobile access.
- Search or browse products
- See size, colour and stock
- Locate it on the store map
- Request assistance on demand
- Leave quick feedback
02 — Staff
Know who needs help — and why.
A simple staff interface answers one question: who needs me, where, and for what? No analytics overload, just clear requests.
- See active requests
- Customer location and need
- Clear priority order
- Accept and resolve in one tap
- Fewer unnecessary interruptions
03 — Retailer
See what happens before the sale.
Anonymous, aggregated interactions become operational insight: demand, friction and customer intent that transaction data can’t capture.
- Searches and unmet demand
- Response times and workload
- Demand by store zone
- Peak periods
- Friction points across stores
Future capability · Recommendations and alternatives are proposed capabilities explored in the thesis. Production AI, live inventory and retailer integrations still require implementation and evaluation.
Proposed AI layer
Understand the question. Connect the right information.
The thesis proposes an AI agent that interprets product questions, connects them to store information and supports alternatives. The aim is to structure customer needs into useful requests and signals — not to replace staff judgment.
- 01
Customer question
- 02
Retailer product and stock information
- 03
Relevant answer or human assistance
- 04
Structured interaction signal
Live AI and reliable retailer data still need engineering and evaluation.
Data and dependencies
Useful answers start with reliable data.
Product catalogues, size and colour availability, stock updates and store maps would come from the retailer. Stale or unknown availability must be made clear. RFID was raised in interviews as a possible reliability enabler — no RFID or POS integration is live.
The intended design uses voluntary interactions and aggregated operational signals, not facial recognition or individual shopper tracking. Security, access, retention and legal compliance need validation before deployment.
Access points
One system. Different ways in.
The interface should fit the store, not define the product.
Prototype interface
In-store screen
A shared point for product information and assistance.
Proposed access
QR / smartphone
A mobile entry point to reduce fixed hardware and queue friction.
Future capability
Retailer app
An integration route to explore with a retailer.
What could OnSight reveal inside your store?
Explore how OnSight could help connect customer needs, staff response and store intelligence in a real retail environment.