We Skipped the Engineering Queue and Built the Tool Ourselves
We replaced months of engineering and a 10-page doc with a single analyst and Google Gemini.
Is getting on the engineering backlog slowing your team down? Is writing a business case taking longer than just building the thing? Feeling pressure to use AI but not sure where to start? WE SEE YOU. And we can help.
A Brilliant Muse program manager used AI and working knowledge of HTML to prototype, iterate, and deploy a working internal tool — in weeks, without a business review doc, without needing to get in the engineering team’s queue, and without a full dev team.
75%
reduction in time to move from first prototype to working internal tool, freeing up time and resources
5X
faster iteration — feature changes turned around in one business day vs. the standard multi-step engineering cycle
$330K
estimated engineering cost savings vs. hiring a 3-person team for 4 months
Highlight Reel
The Challenge
A UX research team needed an incentives calculator - a tool to help them calculate incentive amounts for research studies and determine purchase parity globally. The gap between the current process and ideal state was real: the existing spreadsheet worked, but it was slow and clunky, not the kind of experience a UX organization should be delivering to its own researchers.
The ask from the client team was to build an app. The problem? Engineering was backed up, the backlog was long, and getting prioritized meant writing extensive documentation that could go back and forth for weeks before a single line of code was written.
An early attempt using Google AppScript produced something that looked too much like the spreadsheet it was replacing — same structure, better skin. The client wasn't satisfied. The direction wasn't clear enough. And there wasn't an obvious platform or blueprint to start from.
Rather than cycling through another round of documentation, the Brilliant Muse PgM went back to the drawing board and built.
Starting with existing internal tools as a design reference, the PgM hand-built the first page of a multi-page form experience using basic HTML knowledge. Then, instead of continuing that process six more times, they used AI to do what AI does best: scale and refine content that already exists.
The result was a multi-page, brand-aligned web app — a lightweight internal tool that guides researchers through incentive calculations step by step, pulling from a spreadsheet backend and surfacing the information they need fast.
No third-party CSS. No complex server infrastructure. No dedicated engineering support. Just a working tool that looked and felt like it belonged within the internal tool ecosystem.
Our Solution
What was incredible is that not only did you develop a new workflow, but also how aligned to our brand it was. It felt like it was built close to what an internal tool would have been. It doesn't feel like it was built by another team.
Brilliant Muse Client, UX Program Manager
Under the hood
For the technically curious, here's how it all came together.
The Method & Tools
VIBE CODING
Using AI to iterate on code conversationally — describing what you want, reviewing the output, refining — rather than writing line by line. This powered the rapid prototyping phase and made feature changes possible within a single business day as opposed to lengthy back and forth with rounds of reviews and revisions.
GOOGLE GEMINI CANVAS
An interactive, side-by-side workspace within Google Gemini designed for real-time collaboration on coding and content. Canvas enabled the PgM to build, preview, and refine the prototype in a live environment — collapsing the traditional prototype-review-revise cycle.
HEURISTIC EVALUATIONS
Used Gemini to stress-test the tool against established usability principles — surfacing UX friction points before rebuilding, not after. This brought expert-level evaluation precision to a one-person build process.
The Process
Build first, then let AI scale.
The PgM built page one by hand, establishing the design constraints, brand alignment, and logic architecture. AI was then used to scale that foundation across all six pages and introduce complex business logic that would have taken weeks to write manually. This is a key principle: AI refines and expands existing content far more effectively than it creates from scratch.
Keep the stack simple & familiar.
JavaScript-based web app: No database connectivity or server-side scripting required
Google Sheets as the data backend: Familiar to the client, updatable without an engineer
Simple, low-cost hosting: No need to spin up a dedicated server or virtual machine
Run user testing the UXR way.
Once the working prototype existed, the PgM ran structured walkthroughs directly with UX researchers — the end users of the tool. Researchers drove through real scenarios, flagged gaps, and requested features. Changes were implemented within one business day using AI, bypassing the standard cycle of bug submissions, engineering assessment, code review, and deployment.
Expand beyond the original ask.
The tool ultimately delivered more than a spreadsheet replacement. Feature requests from UXR walkthroughs added capabilities the spreadsheet never had: comms planning support, data exports, and guided user journeys that helped researchers move through studies faster.
What really shines is the subject matter expertise. We can have more discussions about user experience and less about documentation. We're less disoriented about what's technically feasible and more focused on how to solve the problem.
Brilliant Muse Client, UX Program Manager
Why it Worked
Domain expertise first. Our Program Manager had deep Research Operations context so that resulted in no ramp-up time and no guesswork about what researchers actually need.
AI as an accelerant, not a crutch. Vibe coding powered rapid iteration. Real technical judgment decided when to use it and when to put it down. The prototype was the starting line for production-quality work, not the finish line.
Prototype as communication. Showing stakeholders a working tool compressed weeks of alignment cycles into days of real user feedback. Less documentation, more value.
Scrappy by design. The tool fits the actual need: five to ten daily users, a spreadsheet backend, brand-aligned user interface. It doesn't pretend to be something it's not, but it gets the job done