Scout

Our own product: a personal shopper that waits

Scout is a product we designed without a client: an AI personal shopping app for iPhone. People tell it what they need, and it researches every store, explains one clear pick, waits for the right price and only buys inside the limits they set. We sell it as a Figma kit for teams building shopping assistants, price trackers, comparison tools and order experiences.

Shopping is where an AI assistant is easiest to distrust. It spends real money, it can be wrong about what someone wanted, and it is never obvious whose interest a recommendation serves. So the design question was not how clever Scout can be, but how much of its working it shows.

Ask in plain words

A request is a sentence: headphones for travel, under $300, great noise cancellation. Scout answers with what it understood, laid out as chips the person can tap to change, budget, use, priority, the devices they own, and asks one follow up question, here about the length of their usual flights. Before it starts, it says what it will check: prices and stock at six retailers, thousands of reviews, and weight and battery for long flights.

Research you can see, and one pick you can question

The research happens on screen. Each step is listed as it runs, from understanding the brief to comparing prices and reading reviews, and the products ruled out are counted by reason: too heavy for long flights, weak noise cancelling in reviews, over budget even on sale, out of stock everywhere. Three finalists remain, and then one pick.

The pick says plainly where it breaks the brief. In the example it is $50 over budget, and Scout explains why it still recommends it and that it is below its usual price today. The person can watch it for a lower price or ask why not the alternative, so the recommendation is a starting point for a conversation rather than a verdict.

Watch the price, buy at the right time

Buying has three levels of trust. Watch tells the person when the price is reached and buys nothing. Hold adds it to the cart and asks first. Buy for me buys within the limits, with a window to cancel. A target price comes with an outlook of when Scout expects it, and the watches list shows how much waiting has saved over the year.

Buy for me writes every limit out before it is switched on: the most Scout can spend including tax, the price to buy at, the exact product, the retailers, the address, the card, the cancel window and the date the watch ends. When it buys, the alert says what it paid, what it saved and until when the order can be cancelled.

Every state, designed

A shopping app spends a lot of its life waiting on networks and stores, so the states that are not the happy path were designed as carefully as the ones that are: loading, offline, empty, no results, a confirmation before anything destructive such as cancelling an order, and success.

Same system, new surface

The file includes a landing page of 11 sections built from the app’s own tokens and components, so the site that sells the product looks like the product. It is the same principle we follow when we design a product and the site that sells it together.

Ink and one blue

The brand is deliberately quiet: ink, paper and one blue, with colour used only where it carries meaning. Blue belongs to Scout itself, green appears only for savings and amber only when something is urgent. Type is Geist throughout, with 64 icons drawn in one stroke style.

The 61 screens run across 13 flows at 430 by 932, built from 36 components with 105 variants on three variable collections for primitives, colour and layout. One summary card powers ten screens, which is what keeps a mobile kit consistent as it grows. The kit is on our products page.