Our UX Co Pilot

The intersection of UX design and AI is so new that the best practices don't exist yet. We're building infrastructure to discover them—and open access to anyone who wants it.

UX has only been a recognised discipline for about ten years. AI tools are even newer—we're watching the adoption curve spike in real time. The overlap between these fields is basically uncharted territory.

At Willard, we have a specific framework for how startups should approach UX: clarity before creativity, experimentation over perfection, teaching founders while building with them. It works. But not everyone can afford agency services to access it.

So we're building a Figma UX co-pilot—a plugin that brings Willard's framework directly into the design tool, available to anyone who needs it.

The problem with most AI design tools

Most AI tools for designers generate variations, suggest copy, or auto-arrange layouts. They're productivity boosters—useful, but shallow.

They don't teach good UX thinking. They don't surface patterns based on evidence. They don't help founders understand why one approach works better than another.

We can't build a tool that teaches our framework by prompt-engineering a language model. We need data infrastructure that captures real UX patterns, tests them with users, and surfaces insights at the right moment.

Building the repository first

Before we can build the co-pilot, we need a robust UX repository. We're assembling it from three data sources:

A web crawler agent that visits sites, documents user journeys, takes screenshots, and structures the flows into a searchable database. Think Mobbin, but built to feed recommendation engines.

We're early—still refining the agent's workflow. Once stable, we'll document flows across the same breadth Mobbin covers.

An interview agent that books calls, walks users through different website journeys, and captures qualitative feedback. This adds the missing layer most repositories lack: why patterns work or fail for real people.

Design heuristics and benchmarking data layered on top—guidelines, performance metrics, usability principles that contextualize the patterns we collect.

The recommendation engine that connects it all

Once we have the repository, we build a recommendation engine that combines documented flows, interview insights, and heuristics into contextual guidance.

When a designer works on a checkout flow, the engine doesn't return generic best practices. It surfaces patterns that worked in comparable contexts, explains why they succeeded, and flags where user feedback revealed friction.

This becomes the backbone of the Figma plugin—the infrastructure that makes the co-pilot genuinely useful instead of just generative.

Opening the doors wider

We're not building this to keep it proprietary. The more people who can access good UX thinking combined with AI tools, the better.

Willard's framework works because it reduces cognitive load for founders, integrates UX with product strategy, and commits to iteration beyond handoff. Those principles should be accessible whether you can hire an agency or not.

The Figma co-pilot makes that possible. But the real work is building the systems underneath—the repository, the agents, the engine—that let us deliver clarity instead of just pixels.

We're documenting the build as we go. The infrastructure phase is messy, unglamorous, and essential. The plugin is just the interface for everything happening below the surface.