Technologist Uses LLM Workflow to Manage Mother's Stage 4 Cancer Care

A 34-year-old technologist deployed a large language model workflow in early 2026 to coordinate care for his mother's Stage 4 duodenal adenocarcinoma, demonstrating how generative AI can augment caregiver oversight in complex clinical settings.

Pratik Desai built the system to ingest daily exports from Epic Systems—the dominant U.S. electronic health record platform—into NotebookLM and Claude. The workflow helped identify CAT-scan misdiagnoses and detect emergencies, supporting three interventions that likely extended his mother's life, according to a first-person account published by Business Insider.

The case illustrates practical application of generative models outside institutional settings. Desai's approach bypassed hospital IT infrastructure, instead relying on patient-accessible data exports and consumer-facing AI tools to create a parallel layer of clinical analysis.

(The account represents a single anecdotal case study and has not been independently verified through peer review or clinical trial methodology. Epic Systems did not comment on data export workflows or third-party AI integration.)

The workflow's success hinges on access to structured clinical data. Epic, which serves more than half of U.S. hospital beds, offers patient portals that export records in standardized formats. Desai's system leveraged this interoperability to feed longitudinal data into models trained on broad medical corpora, enabling pattern recognition across lab results, imaging reports, and medication schedules.

Clinical AI adoption has historically concentrated in radiology and pathology, where image analysis offers clear efficiency gains. Caregiver-driven workflows represent a different vector: distributed, ad hoc systems built by technically skilled family members navigating fragmented care. The approach raises questions about liability, clinical validation, and equity of access as AI tools proliferate outside formal healthcare channels.