Agentic AI Transforms Healthcare
Health systems expand the use of agentic AI to manage complex clinical and administrative workflows.
Anonymous
Key Points
1
At St. Luke’s, AI agents addressed specific tasks including eligibility verification, prescription routing, appointment scheduling and documentation. Frequent transfers between departments had lengthened patient interactions, while triage nurses were spending time on clerical work instead of direct medical assessment—contributing to staff fatigue and longer waits.
2
UW Health has broadened AI oversight beyond clinical leaders to include human resources and other administrative functions, and it has increased the frequency of governance reviews as generative-AI use has expanded. Frank Liao said, “Governance actually allowed us to go faster,” describing the process as a trusted “paved road” for adoption.
3
Revenue-cycle leaders warned that confusing automation with AI can lead health systems to overstate results and underinvest in more advanced tools. Courtney McNamee of Altru Health System said, “AI is really good at finding a needle in a haystack, but you still need a person to help fix it or do the work that needs to be done.”
4
Training for agent development is moving beyond chatbot prompting to production-oriented tools and deployment skills, including retrieval-augmented generation, LangChain, LangGraph, CrewAI, vector databases such as Pinecone, and automation platforms including n8n, Zapier and Make.
5
AI’s potential clinical value includes analyzing X-rays, CT scans and MRI scans to identify patterns that may support earlier cancer detection, while analyzing patient histories and outcomes could help clinicians select more individualized treatments and reduce ineffective care.
Health systems are moving beyond one-off AI experiments toward agentic systems that handle multistep workflows on their own. St. Luke's Hospital deployed AI agents to manage roughly 400,000 monthly patient interactions, potentially saving about 100,000 labor hours and more than $3.5 million in monthly direct labor costs Becker's Hospital Review. The shift is forcing health systems to rethink governance, workforce training, and how they measure AI's actual value.
Unlike simple chatbots, these agentic systems reason through multistep tasks—eligibility checks, prescription routing, appointment scheduling—and route work between departments without human hand-offs at every step. But success hinges on governance, clear accountability, and honest assessment of what AI can and cannot do.
St. Luke's Cuts Labor Costs With Autonomous Call Routing
St. Luke's deployed AI agents to handle routine patient calls in its call center. The system performs intent classification—understanding whether a caller needs eligibility verification, prescription help, or an appointment—and routes the call without forcing transfers between departments Becker's Hospital Review. Previously, frequent transfers lengthened interactions and forced triage nurses to spend time on clerical work instead of direct medical assessment.
The impact is substantial: 400,000 monthly interactions, 100,000 labor hours saved monthly, and direct labor cost reductions exceeding $3.5 million per month. This allows nurses to focus on higher-value patient care and reduces staff fatigue from administrative bottlenecks Becker's Hospital Review.
Governance Becomes the 'Paved Road' for Safe Adoption
UW Health reviewed more than 100 new AI features in a single platform upgrade and expanded governance beyond clinical leaders to include human resources and administrative functions. Frank Liao, Senior Director of Digital Health and Emerging Technologies, said governance "actually allowed us to go faster," describing centralized oversight as a trusted "paved road" for adoption Becker's Hospital Review. The system classifies features by risk level and increases review frequency as generative AI use grows.
Liao emphasized that each solution must have clear departmental accountability: "Every solution…has to live in some part of the business that's accountable for the outcomes" Becker's Hospital Review. This approach prevents silos where different departments deploy AI tools without shared visibility or responsibility for failures.
Revenue Leaders Warn Against Confusing Automation With AI
Courtney McNamee, Director of Revenue Cycle at Altru Health System, cautioned that health systems often mislabel basic automation as advanced AI, inflating claimed results and delaying investment in true pattern-recognition tools. Becker's Hospital Review She noted: "AI is really good at finding a needle in a haystack, but you still need a person to help fix it or do the work that needs to be done."
This distinction matters because claims scrubbing and rule-based routing deliver modest savings, while AI that analyzes patterns can unlock deeper value. Revenue-cycle leaders argue that conflating the two leads health systems to overestimate wins and underinvest in transformative tools Becker's Hospital Review.
Workforce Demand Shifts Toward Agent Orchestration Skills
Training programs are moving beyond chatbot prompting to production-oriented tools for building and deploying AI agents. Demand is rising for expertise in retrieval-augmented generation (RAG), orchestration frameworks like LangChain and LangGraph, vector databases such as Pinecone, and automation platforms including n8n, Zapier, and Make Becker's Hospital Review. These skills allow teams to build agents that fetch data, reason through workflows, and integrate with health systems' electronic health records.
Health systems are investing in internal training and hiring to reduce reliance on vendors and consultants. The shift signals that agentic AI is moving from pilot projects to core operations, requiring workforce development that matches technical complexity Becker's Hospital Review.
Clinical Opportunities Face Safety and Privacy Hurdles
AI agents show promise in analyzing medical images—X-rays, CT scans, MRIs—to spot patterns that may support earlier cancer detection. They can also review patient histories and outcomes to help clinicians select more personalized treatments and reduce ineffective care Becker's Hospital Review. These applications could accelerate drug development and improve patient outcomes at scale.
However, clinical deployment remains constrained by strict requirements around data privacy, system reliability, safety, and mandatory human oversight. Health systems must prove that AI outputs are trustworthy and that humans remain meaningfully involved in high-stakes decisions Becker's Hospital Review.
Sentiment
Mostly NegativeNEGATIVENEUTRALPOSITIVE
Publishers
26
Articles
7
Reach
33
A
AnonymousAuthor
