Future Bet: AI Agents

Vision & Goals 

We see agentic AI as covering a whole range of projects, including next-generation robotic process automation (RPA), bots, automated agents and intelligent automation.

Our Solution 

At Houston Methodist, we have been employing agentic AI in scheduling, revenue cycle and hospital follow-up calls for a while. In 2025, we had 7.6 million total calls to manage, compromising inbound inquiries, outbound scheduling, quality assurance and more. With our automation services in place, our patients reported a 95% satisfaction score for ease of scheduling, and we had a $9.5 million return on investment (ROI) due to an increase in completed MRIs, CTs and mammograms. We also saw a 25 to 50% reduction in administrative costs. 

While administrative functions are quickly mobilizing and adoption is seeing success, clinical use cases are moving more slowly. Adoption of agentic AI to replace physician calls or nursing questions is immature and has not moved the needle much on nurse staffing. One area we will stress is the success of text-based care pathways. For seven years, we have used more than 200 text-based care pathways and have witnessed more than 80% current and repeat usage. We anticipate that we will expand and combine our successful text-based pathways with new agentic approaches.

Next Steps

In the next three years, we expect AI agents to support most of our key operations. This work is already underway and spans scheduling, registration, consents, revenue cycle, prior authorization and clinical support in physician offices for tasks such as medication refills or answering basic questions. Our key takeaway is agentic AI is new and can be costly if not used appropriately. That’s why we are deliberate about how, where and when we deploy it at Houston Methodist. With thoughtful implementation, we anticipate significant efficiency gains across many areas while continuing to deliver exceptional experiences for our patients and clinicians.