Discovery
AI and technology
AI-powered ideation
Our clients had a need to power their ideation process with AI. We built OptiPilot AI, a tool scans that a webpage and consumes various data formats to generate hypotheses for A/B tests.
Back
Case study optipilot-ai
Supercharging ideation with a custom built AI-powered tool for our clients
Technologies
The tech stack we used to build this project
Open AI
Vue
TypeScript
Node.js
Docker
Tailwind
Project summary
We built OptiPilot to help our clients radically speed up and improve the process of generating high-quality experiment ideas. Designed as a smart AI assistant for experimentation teams, OptiPilot uses a custom-trained AI model to turn web analytics, goals, and context into structured test hypotheses — all delivered through a lightweight Chrome Extension.
The problem
For many of our clients, coming up with new and effective A/B test ideas is one of the most time-consuming parts of their optimisation work. Ideation is often ad hoc, inconsistent, and lacks the benefit of accumulated experimentation knowledge. While general-purpose AI tools like ChatGPT offer some support, they don't have the domain-specific understanding or formatting to generate credible test specs out of the box. We saw an opportunity to build something better — a custom tool that could generate tailored test ideas informed by experimentation best practices and client-specific context.
The challenge
We needed to create a tool that was simple to use but powerful in output — capable of understanding a page's goals and generating real test specs, not just vague ideas. The extension had to be self-contained, privacy-conscious, and easy to run without onboarding friction. We also had to manage Chrome Extension limitations, handle OpenAI's streamed API outputs via a secure proxy, and ensure all outputs could be turned into clean, client-ready documentation. The user interface had to remain intuitive and fast, despite the technical complexity under the hood.
The services
We designed and delivered the OptiPilot MVP as a Chrome Extension, using Vue.js on the frontend and Node.js for a backend proxy service that interacts with OpenAI. The extension collects qualitative and quantitative data from users, captures a screenshot of the relevant page, and sends this to our AI system for idea generation.
Related case studies
Discover more of our work
Related case studies
Speed to market
Pet care
Delivering a veterinary mobile app at speed
Previously struggling to gain traction, we took over The Net Vet app and delivered a working MVP in 3 months, transforming a stalled project into a successful launch before handing back over to the team
Supercharge innovation and advance without limits.
Collaborate with our team and hear back within 24 hours
Let's collaborate