I drove Weedmaps' organizational adoption of AI-assisted design and design-to-code workflows — evaluating tools against the risk of low-quality "AI slop," landing on Figma Make, Claude, and Builder IO Fusion, and pairing the rollout with a parallel design-system modernization effort that made AI-generated output production-safe. I used the resulting workflow to take a search redesign, an AI chatbot prototype, and a desktop navigation overhaul from concept to executive-approved initiative within 30 days, earning same-day CEO sign-off on a full desktop overhaul.

The problem. Most design orgs were still experimenting with AI tools rather than operationalizing them — and the obvious risk of moving fast was more "AI slop": faster output that didn't actually serve customers or the business. Our design process also relied on designers producing artifacts that still needed separate engineering implementation, creating a handoff gap AI could either close or widen.

Why me. I'd been an early adopter of ChatGPT before Weedmaps, going as far as building my own GPT as a personal experiment, and working with Claude to produce my own iOS app — giving me both hands-on fluency and an early, clear-eyed view of where AI tools actually helped versus where they just generated noise.

The decision. Rather than adopt tools reactively, I ran a deliberate evaluation against the AI-slop risk and landed on a specific stack: Figma Make and Claude for design-to-code prototyping, and Builder IO Fusion for close-to-code implementation and lower token cost via its visual editor. Critically, I identified that none of this would be safe to scale without a design system that was fully coded and connected — so I ran the AI rollout and a parallel design-system modernization effort together, partnering with the VP of Engineering to install a dedicated design systems engineering lead (a senior staff engineer and former creative director) who built the CI/CD pipelines and MCP connections that made the system usable by AI tools.

Leading the team through it. The reaction inside the team was genuinely split — some designers were unsettled by "design is dead" industry narratives; others saw a chance to amplify their core value. I managed this with a deliberate cultural practice: weekly affirmation that experimentation was safe, that failure was expected, and that any win — big or small — got celebrated. My Senior Staff Designer, Dennis O'Neal, and I moved fastest and used that momentum to show the rest of the team what was actually possible, rather than mandating adoption top-down. I also personally drove procurement — securing Claude seats for the full team and the Builder IO Fusion license.

The proof. Within roughly 30 days, the new workflow produced: a reimagined search experience (built on jobs-to-be-done and customer journey work, visualizing real-time product availability by dispensary location — an experience without a direct competitor precedent); an AI-powered "Budbot" virtual budtender chatbot working on real data, which Dennis prototyped under my direction with the ML team; and a desktop information-architecture and visual-system overhaul, including a new radial add-to-cart icon that drove a matching-radius redesign across every card component. When I presented this work, the CTO was excited enough to bring it to the CEO the same day — who responded "how can I have it all" and authorized a full desktop navigation and design overhaul, with spillover into app and mobile web.

The outcome. With the design system rebuilt and the tool stack in place, we established a gated model: designers could branch directly from production, build with AI assistance, and ship front-end code validated by engineering before merge. This closed the drift problem — designs and production no longer diverged, since designers were now working from the same source of record as engineering instead of reconciling stale files. It also changed the economics: by absorbing front-end implementation work directly, we freed native and front-end engineers to focus on backend and platform work, and I estimated the shift saved Weedmaps north of $300K annually in front-end dev time (scoped to front-end only, excluding middle- and backend work). The net effect was a faster ideation-to-ship cycle — designers effectively became an extension of the front-end team, and every project we pursued through this model shipped to merge without error, in full fidelity, on live production data.