To open the second day of READY 2025, Scott Gnau announced READY Games winners: Control Robotics for Most Innovative and LEAD North for Most Likely to Use. Among other awardees, Red Salute and Stanford were recognized within the Healthcare Providers category, while Cigna received accolades for Health Plans.
Following the awards, there was a presentation on IRIS. The presenters emphasized how IRIS is timeless. Built with a core data engine for rapid storage of diverse data, capabilities for transactions, analytics, cloud-centric and Generative AI interoperability, and extreme scalability, IRIS facilitates the management of various data types, transforms data into any format, moves processing to the data source (thereby reducing costly data movement), requires a smaller hardware footprint (less CPU, less cooling, etc.), optimizes long-term costs through centralized data maintenance, and minimizes data duplication, ultimately dismantling silos and enhancing organizational efficiency.
In 2025, IRIS expanded its data format support, implemented performance improvements, and introduced Data Fabric Studio with Retrieval-Augmented Generation (RAG) capability. Pete Lesperance from Epic highlighted that their technology supports 305 million patient lives, with MyChart serving as their patient portal. On Epic’s model, all users access the same core patient record, ensuring data consistency across providers. IRIS supports large-scale customers by managing complex, multi-dimensional data and guaranteeing consistency throughout the supply chain, particularly for patient records in extensive health systems.
Real-world scaling via Epic Cloud has so far enabled four customers to manage three times the normal volume of messages. Epic is integrating capabilities such as vector search and AI, underscoring foundational developer-developer interaction with InterSystems. A dedicated testing center in Minneapolis rigorously evaluates new IRIS releases. Performance engineering teams and other relevant groups convene bi-weekly for information exchange and experimental builds, with quarterly project reviews and semi-annual performance and availability meetings held in Verona, Wisconsin, and Boston. These collaborations determine the strategic focus for the upcoming year, aiming for a 15-20% GRefs per second, per core improvement in core efficiency and a 25% improvement in user response time for database users.
Epic demonstrates leadership in healthcare technology by sponsoring hack-a-thons and promoting flexibility with emerging technologies like AI. Their ongoing work includes ECP manageability, security/encryption for regulatory compliance, and accelerators such as CSL 3.0 and shared memory. The latest IRIS version available to Epic customers is 2024, with an 84% adoption rate of IRIS 2024.1 or newer. Epic provides specific platform guidance, characterized as “Midwest-nice, but aggressive with upgrades,” serving as a model for partnership.
Following the IRIS presentation, Gokhan Uluderya, Head of Product, Data Platforms, discussed the profound reliance on data in the AI era. He acknowledged the pervasive hype surrounding AI, but affirmed that AI’s tangible value is undeniable, and that investing in it will pay off. He referenced Gartner’s survey, indicating that, today, however, over 30% of Generative AI projects are abandoned after the Proof of Concept phase, and half require eight months to transition from prototype to production due to challenges such as a lack of AI-ready data (leading to hallucinations), complexity, high cost, and a deficit of trust.
An effective AI strategy begins with a robust data strategy, as data serves as the foundational fuel for the AI engine. Key innovations at InterSystems include: a real-time data fabric for comprehensive data, transactions, and analytics; out-of-the-box AI, Business Intelligence, reporting, and advanced analytics powered by vector search and RAG; orchestration of human and agentic workflows; and common governance, observability, and control for responsible data and AI.
InterSystems endeavors to circumvent the “Trough of Disillusionment” that models predict most companies go through when new technology is introduced by offering automated vector embeddings, foreign tables with advanced query processing, row-to-columnar conversion with enhanced mixed-query performance, improved database management, table partitioning, RAG-in-the-box functionality, AI Agents for data, and an agentic experience for end users. Developer experience enhancements encompass VS Code-first development, Jupyter Notebook Experience within VS Code, and client drivers available in public repositories.
Jeff Fried elaborated on integrated vector search, noting its impressive adoption since its release one year prior. He demonstrated the distinction between using outdated data versus RAG for current information. He also mentioned GraphRAG for thematic exploration and its construction with IRIS, citing AGIMERO (a Guided Agentic Procurement example developed in 36 hours) and IPA (Complex Data Analysis with IRIS Vector Search via BioStrand LENS), which utilizes 440 million vectors, as illustrative examples.
The next presentation was on the InterSystems Data Fabric Studio, which offers domain-specific solutions with pre-built connectors. A new Health Module for Data Fabric Studio was announced. InterSystems offers a private Early Access Program that includes datasets, research support, data models, FHIR support, specialized dashboards, and AI Insights. A demonstration showcased the process of adding schemas from a catalog, applying transforms to rectify incoming data errors, promoting to a target data model, automapping, and connecting to Business Intelligence tools. This type of functionality enables the construction of health data pipelines, the use of BI/analytics tools, and provides business users with secure data access.
Scott Gnau and Don Woodlock then discussed the evolution of the partner program and the vision for healthcare solutions. Sean Kennedy, Head of HealthShare Product, entertained the audience with InterSystems’ re-telling of the “Wizard of Oz” story (now “Wizard of AI”) to address challenges faced by healthcare CIOs seeking trusted data and Generative AI applications. Kennedy, the main character in the story, was not permitted to enter AI City until he learned more about what is required for AI. Kennedy had to meet a sequence of story characters who would each teach him something important about AI preparedness to help him on his quest to enter AI City.
The first leg of the journey focused on Trusted Health Data (showcasing a lion character in the vein of The Cowardly Lion) with a demonstration of searching for a patient and using an Enterprise Master Patient Index (EMPI) and probabilistic/referential tools to link five disparate records into a single patient record, thereby reducing manual effort.
Actionable Analytics (The Scarecrow character) was illustrated with hypertension, highlighting it as a costly yet manageable public health issue. Dashboards were employed to identify care gaps and potential interventions.
The next stop was to visit an Owl Professor, who demonstrated an AI Assistant capable of answering a clinician’s questions about patient medications, laboratory results (categorized), and conditions (with SNOMED/ICD-10 codes). The system settings allow for defining roles, constructing prompts, and controlling data element access for Large Language Models (LLMs).
After saying goodbye to the Owl Professor, Kennedy encountered the Tinman, who spoke to Sean about how AI could significantly reduce time spent on prior authorizations, a known bottleneck in healthcare. The system could scan and ingest forms, generate FHIR-compliant questionnaires for sharing between payers and providers using Generative AI, and pre-populate forms with existing Electronic Health Record (EHR) data, ensuring compliance and readiness for automation.
Kennedy finally gained admittance to AI City. After being discovered behind an invisible curtain, The Wizard (Don Woodlock) revealed that AI agents actually run AI City, where LLMs can directly schedule appointments, send emails, or invoke APIs, transitioning from “writing to doing.”
Others joined Kennedy and The Wizard on stage, including Jonathan Teich, playing an ER physician (although he is also one in real life). The AI City team demonstrated an example of the AI of the future with an Emergency Department patient presenting with respiratory distress. Twitch used an AI prompt to obtain a summary of the new patient’s chart, focusing on pertinent details such as asthma, smoking habits, and recent travel history. The system recorded physician dialogue and initiated actions, with specialized agents collaborating seamlessly with other care teams and departments. The demonstration progressed to humanized AI Agent Avatars providing verbal recommendations to the physician, obtaining permissions to take specific actions on the provider’s behalf for the patient, and coordinating with other agents (e.g., pharmacy consults for sepsis, admission plans including bed booking and ward notification).
The agentic AI workflow reduced what would have been more than ten minutes of work for the provider to mere seconds, with necessary internal documents generated and stored based on historical records during the patient encounter. The avatars and speech-to-text/text-to-speech functionalities can be disengaged, depending on provider or health system preferences. A patient support avatar (an AI agent used by the patient to navigate their care) demonstrated its ability on stage to handle multiple languages for a given patient and respond to inquiries based on available data, schedules, etc., showing examples of patient-centered AI.
The overarching message from the AI City story was that customers using InterSystems technologies today are–and will be–“ready” for the future.
