How Our Claude Developers Built Clinical Workflow Automation for WizTherapy
TL;DR: Claude developers for clinical workflow automation shipped WizTherapy's intake platform as a 4-week PoC running on Claude 3.5 via Amazon Bedrock. It extracts fields from PDFs, Word docs, and scanned intake forms, then generates structured JSON and UI schemas compatible with React JSON Schema Form. Result: 80% reduction in manual effort and 90% accuracy in schema generation across variable document formats, all under HIPAA.
WizTherapy is a pioneer in neurodivergent client care — a business built on empathy paired with clinical rigour. Behind the scenes, though, their clinicians and staff were burning hours on intake paperwork, especially at the onboarding stage. As patient volume grew, so did the drag of manually processing diverse document formats.
We built the workflow that made that go away.
The Problem
Every new client at WizTherapy fills out an intake form. Those forms arrive as PDFs, Word documents, or scanned images — every submission a different shape.
Clinicians had to manually extract information from each one, standardise it, and re-key it into digital systems.
At the volume WizTherapy was growing to, that broke down.
- Onboarding slowed to a crawl.
- Data entry errors crept into clinical records.
- Client experience suffered on the very first touchpoint with the platform.
- Their React JSON Schema Form–based systems couldn't ingest unstructured intake, so integration stayed limited.
WizTherapy needed a clinical workflow automation solution that could handle variable document formats and convert them into structured, editable schemas — without breaking HIPAA.
The Solution
Our Claude developers delivered a 4-week PoC purpose-built for clinical workflow automation. The platform handles document ingestion, AI-powered data extraction, and automatic schema creation, all while preserving transparency, security, and clinician ease-of-use.
Multi-Format Intake Ingestion
Clinical teams upload intake forms in PDF, Word, or image format via a secure web UI or API. AWS Lambda and Amazon Textract handle the initial parsing pipeline.
LLM-Based Entity Extraction and Schema Creation
Claude 3.5 on Amazon Bedrock identifies and extracts form fields using AI and NLP techniques. It generates structured JSON and UI schemas compatible with React JSON Schema Form, removing manual mapping and enabling dynamic form rendering downstream.
Interactive Schema Preview and Editing
A lightweight ReactJS interface lets clinical users:
- Upload and preview extracted schema
- Adjust labels, field types, or validation rules
- Download the final schema for integration
Compliance-First Design
HIPAA compliance was top of mind from day one. All extracted data is logged securely, encrypted using AWS KMS, and monitored through CloudWatch. Discrepancies or incomplete data get flagged for human review before hitting the clinical record.
The Impact
- 80% reduction in manual effort on intake form creation and digitisation
- 90% accuracy in schema generation across varied document formats
- Accelerated clinical onboarding — faster care delivery, lower admin load, no drop in accuracy
- 4-week PoC to working platform. Scope to production-ready pilot.
Key Takeaways for Digital Health and Clinical Teams
Pitfalls to Avoid
- Underestimating how much intake documents vary between clients
- Relying on manual schema conversion at scale
- Skipping audit trails in workflows that touch clinical data
What to Prioritise
- Treat intake forms as structured data assets, not paperwork
- Use AI to accelerate both admin and clinical care workflows
- Adopt React JSON Schema Form (or similar) for dynamic, low-code intake interfaces backed by generated JSON schemas
Build the Same Automation for Your Clinical Practice
If intake paperwork is capping how fast your care team can onboard clients, our Claude developers can scope a similar platform on Claude via Amazon Bedrock in a 4-week PoC.