Inspirata AI Reports Record Second Quarter as Five Academic Cancer Centers Adopt Its Registry Platform

Five contracts, four states, no names

The five centers were described only by geography: Arizona, southeastern Minnesota, southeastern Georgia and two health systems in the Philadelphia region. Each was looking for the same thing, according to the company: one end-to-end registry system that automates reportable-only casefinding and abstraction in real time and can show a return across the oncology service line. The one customer voice in the release is unattributed, quoted as "a senior executive stakeholder from one of the Philadelphia-area health systems," which leaves the buyer-side claims unverifiable outside the company.

Inspirata AI cites no study for its 99% accuracy claim

Inspirata AI describes a patented native-AI platform combining deep search, knowledge bases, inference and machine learning to surface reportable cases as source documents are generated, drawing on direct AP-LIS integration and FHIR-based connectivity to radiology and clinical documentation. It puts accuracy at 99%, counts more than 600 captured data elements including molecular and genetic markers, and says the system applies SEER tumor rules through automated inferencing. No study, audit or registry-accreditation review is cited for either figure. The release also asserts that approaches built on large language models depend on data reaching the EMR first and internal coding after, producing six to eight week delays in harvesting data; that comparison names no competing product.

The product page for the joint offering runs different numbers alongside the same workflow. There, an E-Path Plus pre-abstract is populated with 124 data elements and handed to the ONCOLog registry platform with no manual reentry, and the engine is billed as clinically validated and backed by 20 years of NCI heritage. The 124 and the 600 describe separate stages of one pipeline, and neither is benchmarked against what a registrar abstracts by hand today.

George Cernille on mixing model types

Chief technology officer George Cernille put the design choice this way:

Unlike what most people are doing now, using an LLM for everything, even if it is not the best choice, we use our own proprietary algorithms for fast concept detection and inclusion of complex domain knowledge, combined with machine learning and generative AI where each best applies.

The July 16 rebrand release described the same split, a proprietary NLP engine supplemented by LLM API integrations where appropriate. It named Satish Sanan as chief executive and Jim Hendrickson as co-chief executive of ONCO. The company says it serves registries in the United States, Canada and Australia.

Why This Matters to the APO|APE Reader

Inspirata AI is promising real-time reportable-case detection from pathology reports through AP-LIS integration, a claim about synoptic and narrative report parsing that laboratory directors can test against their own back catalog. The pitch in this announcement is financial as much as clinical, turning the registry from a cost center into support for service-line revenue and patient retention, and that framing will decide who signs the check at the next five institutions. A 99% figure with no denominator, no site count and no error taxonomy is a starting point for a request for proposal, not an answer.