AdventHealth has begun deploying OpenAI's ChatGPT for Healthcare across its hospital system to reduce administrative burdens and streamline clinical workflows. The initiative aims to automate time-intensive documentation and support tasks, allowing care teams to reclaim hours weekly and focus more directly on patient care. This move is expected to improve operational efficiency, expand clinical capacity, and enhance the patient experience.
Operating across nine states and serving millions annually, AdventHealth, like many large health systems, faces tight margins, rising demand, and increasing administrative complexity. This pressure often manifests in daily workflows, such as physician advisors spending about 10 minutes per case on utilization management, involving chart reviews, detail identification, criteria checks, and drafting rationales. This time commitment scales significantly across numerous cases.
Administrative burdens extend beyond clinical roles.
Finance, HR, and IT teams frequently dedicate substantial time to drafting documents, summarizing information, and preparing necessary materials. Leaders describe many departments as operating in a "constant operations mode," limiting capacity for higher-value work. Simultaneously, internal interest in AI was growing, with employees experimenting with chatbots despite formal usage restrictions.
Scaling AI for Enterprise Use
Rob Purinton, AdventHealth's Chief AI Officer, noted that many employees were eager to use AI but unsure how to apply it effectively in their daily jobs. Leadership concluded early that isolated pilot programs would not drive meaningful change, identifying the central challenge as ensuring consistent, safe use across a large workforce. "The hardest part of AI in healthcare is getting humans to use it safely, consistently, and at scale," Purinton stated. "We made a decision early on to treat adoption as the product."
This decision shaped the rollout strategy. Instead of framing AI as automation, leaders presented it as a tool to reduce administrative burden and return time to clinicians and staff. "We don’t talk about AI as automation. We talk about time back," Purinton explained. "If we can take a 10-minute review and compress it meaningfully—while maintaining quality—that’s capacity we can give back to our clinicians."
Measuring Impact and Adoption
AdventHealth also designated adoption as a measurable operational metric.
The organization tracks messages per user per business day, excluding weekends and holidays, to establish a consistent baseline. This metric is monitored and managed like other key performance indicators, with regular reviews of targets and trends. To scale usage, the system relied on domain-based peer groups, such as finance teams collaborating with finance teams, to share prompts, workflows, and best practices relevant to their specific functions.
As AdventHealth transitioned from experimentation to enterprise deployment, leadership prioritized tools meeting healthcare requirements for privacy, governance, and reliability. "We chose OpenAI because we weren’t looking for a demo. We were looking for enterprise infrastructure," Purinton said. "The reasoning capability, the structured outputs, and the governance controls gave us confidence that this wasn’t just productivity software. It was something we could responsibly scale across a health system." The system adopted ChatGPT Enterprise and later ChatGPT for Healthcare, which provides additional safeguards for regulated environments, including data protections and compliance support. Physician advisors now use ChatGPT for Healthcare to generate structured summaries of patient charts, surface relevant clinical details, and draft initial rationales, reducing the time spent assembling information while clinicians retain final judgment. AdventHealth measures impact using system-level data, including timestamps in electronic health records, rather than self-reported estimates. "We prefer measures that are baked right into the process," Purinton added. "We can see exactly how many minutes have improved and whether that change is statistically significant."
COMMENTS