Skål, commute warrior! The RITA — Real-time Information for Transit Agencies — project was built for San Francisco's BART network, and it was mine from first spark to final delivery. I conceived the idea, designed the entire interaction, created the project plan, supervised prompt recording, and produced the marketing materials. Everything but the code.

Initially voice-only but designed from the start for multimodal expansion, RITA provided real-time schedules, fares, and transfer details via live BART APIs — no canned data. Location services tapped into iOS and Android mapping to direct riders to the nearest station with step-by-step directions. Only the Clipper card balance and fare-loading features were simulated; everything else was real-world, real-time.

Customers were leaving — not because they didn't want the service, but because the vacation hold process was so impossibly complex that they couldn't navigate it. A rolling 365-day window governed how many holds a customer could take, for how long, and under what conditions — rules so convoluted that even agents struggled to explain them. Customers gave up and cancelled instead.

I designed a solution that was neither fully chatbot nor fully IVR, but an early example of Agentic AI — a system that could interpret the rules, assess each customer's unique eligibility in real time, and guide them through the process conversationally. Complex, undocumented business rules made simple, and customers who would have walked away stayed.

I led design for an AI-assisted voice-forward customer interface — with a companion chatbot — for a major financial services firm whose clientele included long-established wealth holders with newly-opened or recently-transferred accounts. These were clients with leverage: substantial portfolios, often consolidating from multiple prior institutions, and little patience for systems that couldn't keep up. The design challenge was conveying the status of multiple accounts — active, suspended, pending, in-transfer, frozen, and several variants in between — in a way that was secure, conversational, and easy to navigate by voice. Business rules were extraordinarily complex; requirements shifted constantly through the engagement. The result went live on schedule, and made the complicated sound simple — and natural.

Slàinte Mhath ("SLAN-cha vah") is a traditional Celtic toast meaning "Good health to you" — fitting, as I led the conversational design for a multi-state Medicaid provider services platform pulling forward from legacy DTMF. Providers across multiple states had been navigating button-press menus that predated most of the modern conversational AI field — systems originally built for an earlier era of telephony, layered with decades of incremental fixes. The work: design a voice-forward conversational system that could handle benefit confirmation, claims status, and enrollment verification across state programs that all served the same core functions but manifested them differently. Built on the Kore AI platform with a reusable scaffold framework, the new system let providers anywhere get the answers they needed in natural language — and let designers work consistently across the entire platform regardless of the state they were building for.