The EHR is increasingly becoming the place where health systems encounter, deploy and build artificial intelligence.
Recent moves from Epic, Oracle Health and OpenAI, along with deployments at health systems across the country, show AI moving deeper into the systems clinicians and patients already use rather than operating solely through separate applications.
For CIOs, the shift could change not only how health systems adopt AI but also how they evaluate the growing number of tools competing for a place in clinical workflows.
“AI that isn’t embedded in the EHR is like having a capable assistant who sits outside the room,” Muhammad Siddiqui, CIO of Richmond, Ind.-based Reid Health, told Becker’s in February.
Here are seven recent developments showing how the EHR is becoming a key entry point for healthcare AI:
1. Health systems are building their own AI agents within Epic
Twenty-five health systems participated in Epic’s second Agent Factory Build-a-Thon in September, developing AI agents for clinical and operational workflows.
Projects included agents designed to identify patients at risk of surgery cancellations, review medication refill requests against clinical protocols and patient charts, identify missing critical care documentation and support patient safety investigations.
Epic’s Agent Factory is designed to let organizations build, monitor and evaluate AI agents, Becker’s reported Sept. 15.
Training for an initial group of early adopters is scheduled to begin in October, with broader availability planned for 2027.
2. OpenAI is moving directly into Epic workflows
OpenAI added an Epic EHR integration to ChatGPT for Healthcare in September, with San Francisco-based UCSF Health serving as a pilot partner.
The integration allows authorized clinicians at organizations with supported Epic environments to bring information such as appointment notes, lab results, medications and specialist documentation into ChatGPT or use the technology within Epic workflows.
The integration is read-only and does not write information back to the patient record.
3. Oracle is putting an AI agent in nurses’ EHR workflows
Oracle Health launched a clinical AI agent for nurses in September that is embedded in its EHR.
The tool includes voice-enabled charting, patient information search and acute nursing summaries intended to help nurses more quickly understand a patient’s history and current condition and complete documentation.
The nursing tool follows Oracle’s AI agent for physicians. The company also expanded its Clinical AI Agent in August with automated professional fee coding, AI-powered chart review and clinician-controlled dictation.
4. AI is also moving into the patient portal
The shift is not limited to clinicians.
Sacramento, Calif.-based Sutter Health became the first health system to go live with Epic’s Emmie AI assistant for patients. Indianapolis-based Community Health Network later launched the tool inside MyChart.
Emmie allows patients to ask conversational questions about information already available in their records, including visit notes and test results, and can guide patients toward actions such as scheduling an appointment.
Oracle Health is taking a similar approach. Its new AI-powered patient portal, which became generally available in August, integrates directly with Oracle Health’s EHR. Its embedded AI can answer questions about a patient’s record in plain language and cites the information used to generate its responses.
5. Providence is scaling AI through Epic
Renton, Wash.-based Providence has increasingly used Epic as a distribution point for AI across clinical and administrative workflows.
The 51-hospital system launched 12 Epic AI use cases after an April EHR upgrade, spanning inpatient, ambulatory and revenue cycle workflows.
By July, Providence reported an 84% overall acceptance rate for AI-generated drafts across 14 Epic workflows during the first week after its recent Epic upgrade.
The health system has also told Becker’s that it is looking to consolidate more functionality within Epic when possible rather than relying on separate point solutions.
6. New Epic implementations are arriving with AI already attached
For some health systems, AI is no longer something added years after an EHR installation.
Mullica Hill, N.J.-based Inspira Health launched Epic across all of its facilities July 27 and deployed more than 50 digital tools from Epic’s AI suite and cognitive computing models beginning on the first day.
The tools support functions including documentation, translation, organization of clinical information and predictive analytics.
Inspira offers one example of an EHR implementation launching with AI capabilities already built into the rollout.
7. Health systems now have more embedded AI than they can deploy at once
The growing volume of EHR-based AI is creating another problem for CIOs: deciding what is actually worth turning on.
Clara Lin, MD, vice president and chief medical information officer at Seattle Children’s, recently told Becker’s that Epic is releasing too many AI tools for the organization to pilot each one individually.
Seattle Children’s instead prioritizes the tools most likely to produce meaningful returns before committing resources to them.
Other health IT leaders described a similar tension. Some organizations are increasingly inclined to start with technology already integrated into Epic because of the costs and complexity associated with implementing and maintaining separate tools, while others continue to use third-party products when they believe those tools offer capabilities the EHR vendor does not yet provide.
That distinction matters.
Health systems are not handing their entire AI strategies to EHR vendors. Leaders have told Becker’s that they still expect to use a mix of EHR-native and third-party technologies.
But as AI moves into documentation, chart review, coding, patient communication, nursing workflows and autonomous agents, the EHR is increasingly becoming one of the first places health systems look to put it to work.
