From Disconnected Processes to Unified Field Intelligence

Protagona partnered with the healthcare division of a larger services organization to build a cloud-native Visitation Assistant on AWS — equipping service managers with AI-prepared briefings and voice-captured notes before and after every member visit.

Industry

Healthcare

Teams & Services

Solution Architecture, Data Engineering, AI/ML Integration, DevOps, Engagement Management

Tech & Tools

AWS Lambda, Amazon S3, Amazon Cognito, AWS WAF, Amazon Bedrock, AWS AppSync, Amazon API Gateway, Amazon CloudFront, GraphQL, Terraform, REST APIs, Voice-to-Text, Python

Key Data Points

All four Visitation Assistant use cases — Identify, Preparation, Voice Notes, and AWS infrastructure — scoped, designed, and delivered within a six-week engagement.

Voice recordings converted to structured summaries and prioritized action items through a two-stage AI pipeline built on Amazon Bedrock and AWS Lambda.

Full knowledge transfer completed: the client's internal IT team independently owns, deploys, and operates the Terraform-managed environment from day one.

The Vision

The healthcare division of a larger services organization set out to unify how field teams operate through a next-generation internal platform. At its center: a Visitation Assistant built to give service managers AI-powered preparation and structured note-taking for member visits. This project moved a validated proof-of-concept into fully operational, cloud-native workflows — replacing manual preparation and disconnected processes with a modern AWS-backed Single Page Application.

The Goal

Protagona was engaged to deliver a SPA supporting two core workflows — Visitation Preparation and Voice Notes — backed by serverless compute, LLM-powered summarization, and SAML-based authentication. Success meant service managers could enter member meetings with AI-generated briefings and capture structured notes by voice, with all infrastructure documented and transferred so the internal team could operate the platform independently.

The Challenge

The engagement required building two distinct AI-powered workflows within a single, secure application — each with different data sources, processing patterns, and reliability requirements. The Visitation Preparation workflow needed to pull account-level data, summarize it using an LLM, and export a formatted briefing document — but a direct connection to the source system wasn't available, requiring a CSV-based intermediary and ongoing coordination to keep inputs current.

‍

The Voice Notes workflow added a second layer of complexity: converting service manager recordings into structured notes and action items, with a fallback path to nightly CSV exports when the downstream API was unavailable. Both workflows had to operate reliably while remaining simple enough for an internal team to own and maintain after handoff.

‍

The Solution

Protagona built two AI-powered workflows within a single, secure application. The Visitation Preparation workflow pulls account-level data, summarizes it using Amazon Bedrock, and generates a structured briefing document ready for export before each member visit. Where direct source system connectivity wasn't available, a CSV-based fallback pipeline kept the workflow running regardless of upstream availability.

‍

The Voice Notes workflow follows a two-stage AI pattern: transcribing service manager recordings and transforming raw transcripts into structured notes and prioritized action items, with optional integration into downstream record-keeping systems. All infrastructure was defined in Terraform, and the engagement closed with a live handover session covering the codebase and deployment procedures — confirming full internal team ownership from day one.

‍

AI-Prepared Briefings at Scale

Service managers enter every member visit with LLM-generated briefings drawn from account-level data, replacing manual preparation with structured, exportable documents.

Voice Notes to Structured Summaries

Field recordings are automatically transcribed and converted into prioritized action items through a two-stage AI pipeline, with a reliable fallback when downstream APIs are unavailable.

Internal Team Cloud Ownership

Terraform-defined infrastructure and a live handover session gave the client's IT team full ability to deploy, modify, and operate the platform independently from day one.