UChicago Medicine is putting AI in the front lines to change how patients reach help. Today, human agents answer millions of calls. Many ask for directions, refills, or a basic schedule tweak. The process is full of inefficiencies, delays, and opportunities to improve satisfaction. Using AI, UChicago Medicine is implementing a system that will route those routine moments to AI agents in the first line of response, keeping their human counterparts where more handholding is needed.
“We’re on a mission to transform access and to transform the patient experience,” said Tyler Bauer, Senior Vice President, System Ambulatory Operations at UChicago Medicine at the Salesforce Dreamforce 2025 event.
The Role of Bots and People in Health Care
Today, UChicago Medicine fields more than 3 million calls a year, with over 200 agents supporting roughly 1,500 physicians. Patients phone in to confirm an appointment, request a prescription refill, reach a care team, or ask where to park. Much of that load repeats hour after hour.
“We have about probably more than 50% of the calls that we receive, we feel can be handled by an AI agent,” Bauer shared.
The AI bot rollout begins with low-risk intents such as general inquiries, directions, clinic names and addresses, cancellations, confirmations, and simple scheduling for primary care. Those handoffs free human agents to engage when empathy and judgment matter.
The access center now runs 8 a.m. to 5 p.m., Monday through Friday. With the bots, night and weekend coverage expands too, and callers get a response any time of day.
Next comes the conversational layer. UChicago Medicine has aligned with Salesforce Agentforce for Health to handle tasks like refills, appointment scheduling, and wayfinding with round-the-clock access. Before that shift, agents relied on binders and post-it notes to recall practice rules. Today, a knowledge article pops when a call arrives, tuned to the clinic. UChicago maintains more than 400 such entries.
Bauer frames the change as augmentation. People stay in the loop for the calls that carry risk or nuance.
“It really helps with those redundant tasks that take time away from our live agents, our people, to work on the more complex issues that require that human touch,” explains Bauer.
An Industry Still Operating in the 20th Century
Healthcare still runs on fax. Referrals arrive in free text, often with missing fields. The team intends to centralize inbound referrals to one number, parse the contents with AI, and push a structured work item to the right queue.
“We want to centralize that to one number,” said Bauer. “Using AI to summarize and read what’s in that order and parsed out into different fields to make it easier to read.”
While this might seem to be basic plumbing it has a big impact. Structured referral data shortens the time between a request and a booked slot. It also reduces the number of callbacks, the place where many patients experience sour.
The system goes live near the end of the calendar year. The idea is to start with general inquiries at first and then move to primary-care scheduling. Key things to watch include bot abandonment, track first-contact resolution for refills and appointment changes, while keeping an eye on average speed to respond to requests. In addition, Bauer is keeping a close eye on staff acceptance.
To make sure the technology doesn’t go off the rails or turn into an abandoned proof-of-concept, Bauer says, “make sure you’re thinking about what your problems are first before you look at the shiny technology and communicate really well across teams about what the plan is.”
To help drive adoption, he has brought in new leaders with deep access-center and Salesforce experience. That mix speeds practical decisions on routing rules, knowledge hygiene, and training.
The State of AI Adoption in Health Care
AI adoption in health care is still slow, but there are signs that things are heating up. Kaiser Permanente reports broad deployment of its AI scribes and has ongoing work to guide members to the right care with chat tools.
The U.S. Department of Veterans Affairs has published an AI strategy that names virtual assistants and voicebots for routine questions such as claim status, along with identity-verification and call center efficiency projects. That stance reflects a broader modernization push across benefits and care. Implementation varies by site, and public audits still surface gaps in contact-center performance, which shows how hard the problem remains at national scale.
In London, the North Central London Integrated Care Board piloted an AI receptionist on WhatsApp. Patients can book, reschedule, cancel, and ask questions inside a channel they already use, without a new app.
Risks, limits, and the handoff rule
When it comes to AI in healthcare, part of the reason for the hesitation is a highly regulated environment that is particularly risk averse.
Even outside of delivering patient care, access to customer patient data and patient inquiries requires a safety-first approach. Misrouting a chest-pain message or delaying a chemo visit creates real harm. UChicago’s approach centers on bot-to-human handoffs. If the agent stalls, a person steps in. The team expects to refine prompts and decision trees as seasonal patterns swing, such as shifts during flu spikes.
Bauer accepts the complexity. The data landscape in health care is complicated. Interfaces span CRM, EHR, telephony, payer portals, and referral sources. None of that fits easily. In addition, new tooling helps only if agents trust it, supervisors coach to the metrics, and leaders tell a straight story about roles.
“Folks might have some concerns,” Bauer says. “What does this mean for me? If you talk about the outcome and how this will help you elevate the work that you do, that can be a positive thing.”
However, Bauer insists, “everyone understands this is the direction we need to go. There are multiple connection points and multiple different systems that need to work together to make this happen.”
That last line matters. Call-center jobs burn people out when information hides behind bad data access and difficult to navigate systems. But now with the increasing daily experience with AI, and the drive to modernize healthcare these forces are converging.
Patients want simple paths to care and answers at all hours. Agents drown in repeatable tasks and disjointed knowledge. Bauer’s team describes a change that starts small and grows through consistent, simple wins.
“We’re on that journey right now with our first use case going live at the end of the calendar year. This will transform the experience for our patients and our team members.”
