In a 2023 survey, University of Iowa Health Care found that clinicians repeatedly identified clinical documentation as the single largest barrier to both efficiency and quality patient care. The feedback prompted the health system to actively seek solutions targeting the administrative overload that extended into after-hours work—commonly referred to as pajama time—a persistent pain point among providers.
Dr. James Blum, the health system’s chief health information officer, positioned the adoption of ambient documentation tools as an unavoidable necessity rather than an optional enhancement. “If you haven’t implemented it, you really need to because providers are not going to want to come and practice in your health system if you don’t have a technology like this,” Blum says. The demand for ambient AI tools has surged across healthcare at a pace even faster than the initial rollout of electronic health records.
By 2024, University of Iowa Health Care had already completed a pilot program with Nabla, a specialist in ambient documentation, before deploying the solution across its entire 862-bed academic medical center. A follow-up assessment in 2025 revealed measurable improvements: clinician burnout decreased by 26%, while satisfaction with the electronic health records system climbed by 14%. These findings were later documented in a KLAS Arch Collaborative case study.
Read Also: Healthcare Networks Struggle to Meet Demand
AI market shifts demand new compliance focus
The market for ambient tools has undergone significant transformation since 2022, when Microsoft acquired Nuance Communications, a dominant player in the sector, and ChatGPT became publicly accessible. Today, major health IT vendors like Epic and Oracle Health have embedded generative AI capabilities directly into their platforms. However, compliance remains a critical challenge, shaping how University of Iowa Health Care approaches adoption.
Long before the broader AI surge, the health system had already established a cross-disciplinary team with specialized expertise in evaluating AI models. This group assessed training data, performance benchmarks, and technical specifications, a key advantage when selecting vendors. “If an ambient solution was trained only on a limited data set relevant to one hospital, that was likely not the right product,” Blum explains. The team’s rigorous evaluation process helped the health system avoid early missteps as more sophisticated tools entered the market.
While the initial governance model proved effective, Blum acknowledged it now requires updates. The health system is shifting from evaluating individual tools to adopting a platform-based strategy. “There’s model drift, there’s utilization metrics you need to be looking for, so we’re now looking at re-engineering our governance, which had this lightweight, nimble approach, to become more focused on how we acquire solutions that are platform-based, or how we identify where a tool we already have may address the problem rather than just going out and buying another solution. That’s a work in progress right now,” Blum says. This transition is still unfolding and will take 12 to 18 months to implement across the entire University of Iowa ecosystem.
Strict data rules and seamless integration key
The selection process for the ambient documentation tool was guided by three core requirements. First, the vendor could not use the health system’s proprietary data to train its models, a restriction that immediately limited options. Second, the solution had to be intuitive and adaptable. The chosen product met both criteria, offering a cloud-based system that integrated smoothly with Epic through a minimal HL7 feed. Deployment required only a few days and minimal IT modifications.
Read Also: Edge Computing Accelerates Clinical Decisions at Hospital Bedside
The tool’s accessibility extended beyond clinicians. Physical therapists and other staff members could also utilize it, broadening its application across the organization. Blum highlighted the solution’s simplicity as a decisive factor. “It was so lightweight and easy to install because it’s a cloud-based product that uses essentially a small HL7 feed and has native Epic interchange,” he adds. “We were up and running in a couple of days. We needed to change very little in our IT environment to make it work.”
Beyond documentation: AI for workflows and outreach
As University of Iowa Health Care refines its AI strategy, Blum identified additional opportunities beyond clinical documentation. The health system is exploring AI applications in administrative workflows, including claims processing and patient outreach. He also emphasized vibe coding, a method that allows non-technical employees to build prototypes, as a way to drive innovation without over-reliance on external vendors.
Patient outreach represents another priority for AI optimization. Automated systems could handle routine follow-ups, appointment reminders, and condition-specific health education, allowing staff to concentrate on complex cases. Blum noted that these applications require careful implementation: AI should enhance human interaction rather than replace it, particularly in areas like discharge planning where empathy and clarity are essential. The health system is testing small-scale pilots to assess how AI-driven outreach affects patient engagement and readmission rates.
