Semantic Agent Builder - SAB

The Problem

Clinical knowledge is everywhere. Getting it into an AI agent is the hard part.

Health systems run on protocols — care pathways, outreach workflows, disease management guidelines. That knowledge lives in documents, institutional memory, and the minds of experienced clinicians.

Today, turning that knowledge into a working AI agent requires significant engineering involvement. Every update, every new agent, every protocol change goes through a technical translation layer — slowing deployment and keeping the people who know the most about care one step removed from the tools that deliver it.

SAB removes that barrier.

What is SAB?

A no-code authoring environment for clinical AI — configurable by the people who know care best.

SAB gives clinical informaticists, care coordinators, and operational leads the tools to configure AI agents themselves — defining logic, structuring workflows, and publishing agents without writing code.

Two agent types are currently supported at design time:

Care Pathway Agent — Informaticists upload existing care protocols and SAB's AI extracts and structures them into configurable, stepwise digital workflows. Authors define decision points, link order sets, collaborate in real time, and publish with version control and audit trails.

Patient Outreach Agent — Care teams configure AI agents that handle patient-facing workflows — procedure prep, pre-op checklists, appointment coordination — across phone, text, and chat. The agent's conversation logic, form collection, and escalation paths are all set at design time, in plain language.

My Role

UX lead for the no-code authoring experience.

I lead UX design for SAB's no-code platform as the sole designer — working directly with the Engineering VP and engineering leads to define what we build and how. My focus is the design-time authoring experience: the interaction model for agent configuration, the information architecture of the authoring environment, and the progressive disclosure of complex clinical logic for non-technical users.

With no dedicated PM for the design-time platform, I drive UX requirements and direction for the no-code layer — translating clinical team needs into product decisions, then working with engineering leads to phase and sequence the build.

Health system partners are involved from the research phase — their workflows and mental models directly shape what gets built.

The Agents

Two agent types. Two distinct configuration challenges.

Care Pathway Agent Configure clinical protocols into structured, AI-assisted digital workflows — built by informaticists, governed by institution.

Patient Outreach Agent Configure AI agents that handle patient-facing outreach — procedure prep, form collection, appointment coordination — across phone, text, and chat.

Forward signal: SAB's validation and deployment capabilities are actively in development. These case studies focus on the design-time experience — how clinical teams author and configure agents before they go live.