4.8 stars from 1,500 ratings for an app that invests for you every week

The challenge

What we did

1. 1Design an investing flow people would trust with a weekly budget

   The core interaction is a user naming an amount and then letting the app spend it every week. We designed the sequence around the moment permission is given: what a user reads before the first investment, how much the app explains its reasoning, and what they see the week after money has gone out. Several versions of that sequence were tested against real use rather than argued about internally. The one that held gives the user a standing decision to make once, then visible evidence afterwards. A per-transaction confirmation step would have been the safe design, and it would have quietly undone the automation Benson sells.

2. 2Set up an AI-native workflow, then take on the build

   Design on its own could not keep up with a weekly release. We set up an AI-assisted development workflow with Claude Code and then took on implementation, on iOS features and on the web features embedded inside the app. The marketing site was implemented in Webflow during the same period. The change was in who could finish something: a design no longer waited in a queue to become code. On that workflow the Benson team moved fifteen to thirty tickets a week. The backend stayed outside our scope throughout, which kept the boundary clear and stopped the workflow change from becoming a rewrite.

3. 3One design system across iOS and the embedded web app

   The app is two technologies wearing one face: native iOS screens, and web features running inside the same app. Without a shared system those two drift within weeks and the join becomes visible to the user. We built the design system in Figma and implemented the shared components for the embedded web app in code, so a component existed once as a design and once as a build, and both were ours to keep in step. That is what made a weekly release safe. A new feature could reuse settled parts instead of reopening decisions that had already been made.

4. 4Ship a feature a week, read against the analytics

   Speed only counts if someone checks the result. Benson ran a strict analytics setup, and releases were read against it daily and weekly rather than reviewed at the end of a quarter. That loop decided what got iterated and what got dropped, and it is why the flow around trust went through many versions instead of one. Over almost two years the app held its weekly rhythm and reached 4.8 stars from 1,500 App Store ratings. Pace on its own would have been noise. Pace with measurement is what let the product keep changing without losing the people already using it.