The Hospital CEO Who Replaced His Entire Scheduling Department With AI Has One Year of Data – Here Is What He Found.

There weren’t many journalists or investors in the conference room where Reid presented his findings last month. The audience consisted primarily of hospital administrators, a few anxious department heads, and a few representatives from the tech vendor who had initially sold him the system. It was a Tuesday afternoon in a mid-sized Midwestern city. According to multiple accounts, the atmosphere was somewhere between restrained anxiety and cautious curiosity.

Twelve months prior, Reid had done something that caused the medical community to be equally impressed and alarmed. He completely replaced his fourteen-person scheduling department with an AI platform. Not a hybrid model. There is no part-time coordinator safety net. That was all; the machine was in control.

The choice wasn’t made at random. Regional hospitals’ scheduling departments have long been an odd area of inefficiency, with patients waiting three weeks for appointments that could have been made in five days, phones ringing unanswered, and double bookings. After years of observing that dysfunction, Reid acted quickly when a vendor showed him an AI system that could handle incoming calls, calendar management, patient reminders, and rescheduling at scale. Depending on who you ask, it might be too quick.

It is more difficult than anyone anticipated to summarize what a year’s worth of data actually revealed. The numbers appeared good at first glance. There was a 22% decrease in no-show rates. Wait times for scheduling decreased from eleven days on average to six. With no significant outages, the system managed about 280,000 appointments. Within the first three quarters, the anticipated yearly savings of $1.2 million seemed to be mostly on schedule. On paper, it looks like a success story that someone would present with polished slides to a boardroom.

However, there’s a feeling that something more intricate going on on the ground is hidden by the cleaner numbers. During the first six months, patient satisfaction scores significantly declined, especially among older patients who thought the AI’s phone interface was confusing and impersonal. Many expressed dissatisfaction over the system’s inability to adjust in the same manner as a human coordinator when they had uncommon requests, such as an appointment with a specialist related to a chronic illness or a scheduling conflict due to transportation constraints. Before finding a solution, one patient reportedly called back four times.

For a CEO delivering year-one results, Reid has been rather open about that conflict. He might have realized that it would be more detrimental in the long run to oversell the result rather than admit the system’s shortcomings. In months seven through twelve, patient satisfaction scores did improve, probably as a result of both the AI’s advancement and the patients’ behavioral and expectational adjustments. It’s still unclear if that recovery is a sign of quiet resignation or real progress.

The question of what was lost that doesn’t appear in dashboards is more difficult to quantify, and Reid brought it up during his presentation. The displaced schedulers did more than just handle calendars. They were aware of which patients required more time. Because she worked nights, they recognized the woman who consistently called on Fridays. There is currently no clear metric for the lack of that type of institutional knowledge, and it is difficult to transfer to a training dataset. To be honest, some of it may never be recovered.

Experiments like this one are clearly of interest to the larger healthcare sector, especially in light of hospital CEOs’ public remarks regarding the use of AI to replace radiologists and clinical staff. Although the consequences of missing an appointment are real, they fall under a different category of risk than those of a misread mammogram, making scheduling a relatively contained function. However, as this develops over the course of a year, it’s difficult to ignore how administrators looking for cost savings are beginning to conflate administrative AI with clinical AI.

Delivered quietly in that Tuesday conference room, Reid’s conclusion was basically this: the AI operated, albeit imperfectly, in ways that were worthwhile. He has no intention of reviving the department. However, he does not advocate for other systems to replicate the model in its entirety. It may be the most candid statement regarding AI in healthcare in a long time.