Kuri Saúde × Workflow Intelligence

Scaling hospital workflowsthrough AI automation

Product ManagementAI SystemsOperations Design
Kuri Saúde dashboard

Role

Product Analyst

Timeline

February 2023 – July 2023

Team

1 Product Analyst & 2 Engineers

The Problem

Hospitals receive thousands of pages of pricing tables and negotiation rules from insurers. This data is critical for resolving glosas, but most documents arrive in paper or inconsistent PDFs, forcing analysts to extract information manually. The result: slow reimbursement cycles, high operational costs, and lost revenue.

The Opportunity

Build an AI system that converts unstructured hospital documents into clean, usable data so analysts can review reimbursement cases faster and more accurately.

My Role

Product Strategy

Scoped the problem, aligned constraints, and defined the product requirements for an AI-powered revenue-cycle automation system.

AI Systems & Infrastructure

Researched, selected, and tested OCR/LLM approaches and infrastructure tools. Designed the extraction pipeline and validated it with real hospital data.

Workflow & Prototyping

Mapped operations, designed streamlined review flows, and built interactive prototypes to test usability, performance, and output quality end-to-end.

Feature / AI Pipeline

OCR

digitizes scanned pages

LLM extraction

identifies rules, operators, codes, dates

Normalization

cleans fields, removes noise

Structuring

stores as searchable data

Dashboard

analysts review & act

The Solution

An automated intelligence layer that:

1

Reads thousands of pages in minutes

2

Flags expired or risky rules

3

Structures documents into clean tables

4

Centralizes operator information

5

Powers a new internal dashboard for analysts

Impact

90%

faster document review

80%

lower operational cost

10K+

pages processed in < 5 days

3 + 10

hospitals & insurers using it

Learnings

Infra and models are product decisions.

Choosing OCR tools, LLMs, and the extraction pipeline wasn't a back-end detail. It defined speed, reliability, and what the interface could promise. I learned to treat model and infrastructure choices as core product levers, not technical afterthoughts.

Check other projects