Healthcare delivery is being reimagined through AI agents that work as an extension of your team across critical domains: patient engagement, point of care, and revenue operations. Unlike chatbots or automation tools, these agents reason, plan, and act autonomously. But not all AI agents are created equal. In healthcare, AI agents must perform predictably across millions of patient and clinician interactions while upholding rigorous standards for security, privacy, and ethics. At HIMSS 2026, healthcare providers UC San Diego Health and Pelago shared how they're embracing agentic AI, what they look for when implementing AI agents, how they're leveraging AI-human collaboration to capture efficiencies while staff focus on highly specialized interactions, and the real-world impact they're seeing today.
Amazon Connect Health – New Agentic AI Solution for Healthcare from AWS
During the HIMSS session "Healthcare AI Agents: From Hype to Measurable Impact", AWS presented Amazon Connect Health, a new agentic AI solution that handles patient scheduling, clinical documentation, and medical coding alongside healthcare teams, keeping them informed and in control while delivering proven results across the care continuum.
AI agents are everywhere, but not all are built the same and achieving consistent results at scale is very hard. Large language models (LLMs) that are the backbone of AI agents are generally exceptionally smart, but in healthcare-specific contexts, they need help. To build AI agents for healthcare, we must use advanced post-training optimization techniques such as medical domain understanding and human alignment methods like supervised fine-tuning and reinforcement learning that help focus these models and deepen their knowledge. At AWS, we're building with these advanced approaches along with quality, trust, and safety in mind.
UC San Diego Health: Redefining the Patient Journey
UC San Diego Health is the #1 hospital system in San Diego and the region's only academic medical center. It handles 3.2 million patient interactions annually. UC San Diego Health prioritized improving the patient contact center (PCC) experience, where administrative staff were burdened by high-volume, repetitive manual processes. The team focused on reducing average handle times and enhancing patient interactions through AI. Key capabilities included automatic patient verification and intelligent routing based on caller intent. In addition, real-time reporting dashboards and post-call analytics, including near real-time transcription, enabled the organization to monitor performance metrics such as call wait times, resolution rates, and patient satisfaction in a continuous improvement loop. UC San Diego Health is already seeing the difference: saving 1 minute per call, diverting 630 hours weekly from patient verification to direct patient assistance, and reducing call abandonment rates by 30%—as high as 60% in some departments.
Pelago: Meeting the Surging Need for Substance Use Care
Pelago is the leading specialty substance use care provider, delivering virtual, evidence-based care for alcohol, tobacco, opioid, and cannabis use. Behavioral health is deeply human work. But it is also work that generates enormous administrative overhead—intake assessments, care plan documentation, and ongoing engagement that can overwhelm even the most dedicated care teams. Pelago turned to agentic AI to give them back the time and cognitive bandwidth to focus on the moments that matter most.
After deploying Amazon Connect Health's point-of-care capabilities, Pelago reclaimed hundreds of clinician hours per month and achieved a higher percentage of notes meeting completeness and QA standards, with fewer edits required post-session. This has also translated to better member experiences. Pelago recently achieved its highest NPS to date, supported by care teams who are more present, focused, and engaged during sessions.
Key Takeaways for Healthcare Leaders
Start with the workflow, not the technology. The most successful deployments at UC San Diego Health and Pelago began with a clear understanding of where friction existed and worked backward to identify where AI agents could intervene most effectively.
Trust is earned through AI that cultivates reliability, transparency, and safety. Clinical and operational leaders need to see AI perform predictably across millions of interactions under real-world conditions. And having full evidence traceability where every AI-generated output is mapped back to the specific source data is critical for adoption by care teams.
The goal is augmentation, not replacement. Whether supporting contact center staff at UC San Diego Health or a behavioral health clinician at Pelago, agentic AI works best when it amplifies human judgment rather than substituting for it. The organizations that will lead in this era are those that design AI and human roles together—each doing what they do best.
