By Stephanie Fraser
Conference Correspondent, Answers Media Network
LinkedIn: Stephanie Fraser
Event: Rhapsody Connect 2026
When: October 2026
Where: Boston MA
Rhapsody introduced new AI governance technology at its annual conference in Boston, where healthcare organizations shared lessons on patient identity, interoperability and the operational challenges of AI adoption.
As healthcare organizations expand their use of AI, questions about data quality, security and governance are taking on new urgency. Those challenges were a focus of Rhapsody Connect 2026, held Oct. 20-22 in Boston, where healthcare IT leaders discussed the infrastructure needed to support AI and shared results from interoperability projects.
The conference featured the introduction of Rhapsody Guardian, along with customer presentations examining patient identity, data quality, integration workflows and the growing demands on healthcare IT teams.
Rhapsody Introduces Guardian for AI Governance
Rhapsody CEO Sagnik Bhattacharya opened the conference with a discussion of how AI agents are changing the requirements for healthcare interoperability.
Unlike traditional applications, AI agents can operate continuously, interact with multiple systems and generate significantly more data traffic. They also raise new questions about how organizations authenticate and authorize software agents accessing sensitive information. “How do you manage agent identity? That’s a new thing, and everyone’s trying to figure it out,” Bhattacharya said.
Against that backdrop, Rhapsody announced Guardian, a healthcare-specific API gateway and AI governance solution designed to manage access to healthcare data and services across APIs, applications and AI agents.
According to Rhapsody, Guardian provides centralized authentication, access policies, rate limiting and activity logging across APIs and Model Context Protocol (MCP) connections. It also allows organizations to establish policies governing which data individual applications or agents can access and set limits on their activity and token consumption.
The challenges of governing AI were evident in a presentation from Geisinger, where an influx of requests for new AI applications prompted the health system to expand its review process.
Kim Chaundy, associate vice president at Geisinger and interim president and CEO of Keystone Health Information Exchange, said the organization received 98 AI requests in a single week. It subsequently established two governance committees that now meet four days a week.
“We really never get to a no. It’s not now. We’re not comfortable right now,” Chaundy said, describing the organization’s approach to evaluating applications and addressing potential risks.
Getting Patient Identity Right
Customer presentations also highlighted the importance of accurate patient and member identity as healthcare organizations prepare data for analytics and AI.
John Pyhtila, chief operating officer at Abacus Insights, described a nine-month project with Rhapsody to replace a legacy identity platform supporting more than 13 million member records. The migration required preserving existing matching logic while maintaining daily operations. Through parallel testing and validation, the team reported achieving 99.98% matching accuracy.
“If you can’t link it to the person, and you can’t understand what’s going on with this individual person, you’re not going to be successful,” Pyhtila said.
The project also opened opportunities to incorporate additional data sources, including laboratory information, into future analytics initiatives.
Prisma Health shared its experience using AI-assisted patient matching within its EMPI. Presenters Terri Martin and Kay Hylton Christensen reported reducing duplicate record rates by approximately half after introducing Auto AI, with results remaining stable during six months of cleanup.
The health system also emphasized the importance of human oversight, training data and testing to prevent incorrect patient record matches.
Finding Data Quality Problems Earlier
At Big Sky Care Connect, Montana’s statewide health information exchange, the challenge was identifying problems in data arriving from more than 650 contributing locations.
CIO Drew Hensel described how the organization began using Rhapsody Axon Data Health to evaluate information closer to the point of ingestion rather than relying primarily on downstream analysis. The project included analyzing more than 78 million historical messages. In one example, the team identified approximately 72,000 messages missing medical record numbers.
“Clean data coming in the door is what we want as much as possible,” Hensel said.
The technology also allows staff to investigate problems down to individual message fields and provide participating organizations with more specific feedback about errors.
Interoperability Beyond the Integration Engine
Other customer and partner presentations highlighted efforts to reduce integration complexity and improve access to healthcare information.
Owlet Baby Care discussed its collaboration with Rhapsody to connect data from its FDA-cleared BabySat pulse oximeter to electronic health records through Rhapsody Envoy, while City of Hope shared its approach to identifying risks and improving testing during Corepoint infrastructure upgrades.
Qventus and University of Rochester Medicine discussed using Rhapsody Axon for integration development and troubleshooting, including debugging code, building routes and investigating errors.
Bhattacharya also outlined plans for additional Axon capabilities, including monitoring tools intended to identify integration problems and recommend potential solutions.
Building for What’s Next
Across the discussions, a consistent theme emerged: Healthcare organizations are adopting new AI capabilities while continuing to address longstanding problems involving data quality, patient identity and system integration.
At Geisinger, Chaundy described AI governance as a process that must continually adapt. “It’s constantly tweaking and learning from each other and learning what’s working and what’s not working,” she said.
The experiences shared in Boston illustrated that broader AI adoption will depend not only on new technology, but also on the reliability of the data, infrastructure and oversight supporting it.