Lead Product Designer • Thought Industries • 2024–2025
The outcome: Earned GitHub access mid-2025, shipped an AI product to four paying beta customers, and learned firsthand what changes about design when the most important decisions in your system are invisible to the people using it.
The Challenge
Thought Industries bet that customer education leaders — the people running training programs at B2B companies that already owned an LMS — needed a way to prove the business value of that work. The thesis: if you gave them one weekly synthesis of how the market perceived them (reviews, job postings, competitor content, social signals), they could finally tie training to revenue, retention, and competitive positioning. That was a new ICP, and it shaped everything about how we tested and shipped.
The Process
The arc — from pen-and-paper through wind-down — is told across four articles. Each works on its own; together they’re the full story.
- Part 1 — The brief that changed my job title — the wireframe round where my simplify-first instinct was wrong, and the surprise that reframed the problem.
- Part 2 — The swipe that didn’t work — a Tinder-style UI built to force decisions, and why the gesture wasn’t the problem.
- Part 3 — The integration we couldn’t build (and the one we already had) — pivoting from Salesforce to LMS data, and what changed about design once I was in the codebase.
- Part 4 — Four paying customers and a product that didn’t ship — the honest reckoning: wrong ICP, abandoned core customers, new CEO, wind-down.
Prototypes
The pieces below are the prototypes that carried the biggest decisions in this project — each one led to one of the smaller stories in the arc above.
Embeds landing soon.
The Impact
Four paying beta customers used the product. It didn’t scale, and I’ve written honestly about why in Part 4. The direction wound down as-shipped — but the learnings didn’t. They seeded a second iteration currently in active testing at Thought Industries, one I can’t detail publicly yet.
What matters here is what the work did for me: it made “Design Engineer” a real title instead of an aspiration. GitHub access, shipping directly into the codebase, and eighteen months designing a system whose most important decisions were invisible to the people using it.
Reflection
The question I left cva-01 with was: when the most important decisions in your system are invisible to the people using it, what do you owe them?
Eighteen months of shipping CVA taught me the honest answer. You owe them the shape of the decision, not the machinery. Users don’t want to see the model. They want to see where they can push, what changes if they act, and what the system is willing to admit when it’s uncertain.
Every good moment in CVA was one where we made a hidden decision visible enough to be argued with. Every bad moment was one where the interface pretended the system was sure and the user had to guess whether to trust it. Designing for AI is designing for that seam — where a confident-looking surface meets an uncertain-underneath — and treating it as the primary UX object, not a footnote.
Lead Product Designer • Thought Industries • 2024–2025