Granted Health

Granted is an AI-native healthtech startup helping people save on medical bills and navigate insurance with AI plus human advocate support.

I work across product, design, and development on the core consumer app experience. The challenge is not just making an AI chat feel useful. It is figuring out where an AI-first product should become more structured, where users need explicit controls, and how to keep a human advocate in the loop without making the product feel fragmented.

Product Challenge

Healthcare is a difficult place to make a purely conversational product. Users are often stressed, the source material is confusing, and the product has to collect enough precise information for someone else to act on it later.

That makes the interface problem unusually interesting: the AI should help users understand what is happening and what to do next, but it should not hide important decisions inside a thread of messages.

Chat Plus Structured UI

The core experience starts in chat, where users can describe their bill, upload documents, ask questions, and get guided toward a course of action.

From there, the product needs to introduce structured UI at the moments where precision matters: document intake, insurance linking, HIPAA and consent, payment authorization, financial assistance eligibility, and handoff details for the advocate team.

I have been focused on designing and building that middle layer between free-form conversation and traditional app flows. The goal is to keep the AI experience fluid while giving users clear checkpoints, explicit confirmation, and enough context to trust what is happening.

Human Handoff

Granted is not trying to replace the healthcare advocate. The product has to help the AI and the human team work together.

That means the app needs to collect information in a way that is useful downstream, not just conversationally satisfying in the moment. The handoff from AI to human agent depends on clean documents, known consent state, insurance context, payment authorization where needed, and a clear record of what the user is trying to resolve.

The design work is partly about the user-facing experience and partly about the operational shape underneath it: what does the AI need to ask, what should the UI make explicit, and what does the advocate need in order to move the case forward?

Cross-Platform System

I have also been building and extending the design system for a shared Tamagui codebase across desktop web, mobile web, and native mobile.

The system has to support chat surfaces, dense intake states, structured forms, document review, responsive layouts, and native-feeling mobile flows without turning every surface into a one-off component.

20x Workflow

I have also been helping with internal 20x product-development systems. The idea is that AI-native teams should be able to produce much more output than traditional product teams by changing how work is planned, parallelized, reviewed, and tested.

In practice, that means worktree-based parallel development, drive-by agents for small fixes, AI-assisted design iteration, automated testing, and deeper review loops.

The most interesting thread is design exploration inside real product constraints. Instead of using AI to generate one-off mockups, we have been exploring systems that can search for references, group interface directions, generate multiple design-system-aware iterations, and then move toward UI that can actually fit back into the app.

That matters for this kind of product because the right interface pattern is often not obvious upfront. Some moments should stay conversational. Some need a form. Some need an inline card, a confirmation state, or a more explicit workflow. The value of AI-assisted design is being able to explore that decision space faster while still grounding the output in the product system.

This work is ongoing, so the shape is still evolving as the product and operating system mature.