When I recently called a physician’s office to speak to an actual receptionist about a complex patient I had hoped to refer, I was placed on hold for 15 minutes which then extended to 30 minutes. Eventually, I was transferred—not to a person, but to chatbot and then a voicemail. I left a detailed message and waited for a call back--which never happened.

I won’t name the healthcare system because, increasingly, it rarely matters. The experience has become emblematic of American healthcare: layers of telephone menus, centralized scheduling systems, patient portals, automated reminders and electronic queues standing between one human being and another.

This is separate from the larger debate about AI, and even from the electronic medical record “experiment” I have previously covered for Forbes. The problem is more fundamental: healthcare organizations have spent years removing people from the system in the name of efficiency.

We are now discovering that some of those people were not inefficiencies at all--they were the infrastructure.

When Efficiency Stops Being Efficient

Healthcare has embraced automation for understandable reasons. Labor is expensive, and physician practices face narrow margins. Large healthcare systems are dealing with rising costs, staffing shortages and extraordinary administrative complexity. Software promises to process more calls, schedule more visits, route more messages and reduce the number of employees necessary to keep an office running. For highly predictable tasks, automation can certainly have value and function effectively.

A patient does not necessarily need a human being to receive a reminder about tomorrow’s appointment or to request a routine prescription refill. Online scheduling may be more convenient than waiting on hold. Electronic portals can give patients rapid access to laboratory results, medication lists and portions of their medical record. The crucial mistake is assuming that because automation works for simple transactions, medicine itself can be organized as a series of simple transactions. It simply can’t.

A patient calling a medical office may be trying to explain that her elderly father was discharged from the hospital yesterday but cannot obtain one of his medications. A primary-care physician may need to speak directly with a specialist because a CT finding does not fit neatly into an electronic referral form. A family may be trying to move an oncology appointment forward because a loved one is deteriorating rapidly. These are not merely scheduling problems. They require judgment, context and, sometimes, compassion. Yet we increasingly funnel all three through the same automated architecture.

“Press 1 for appointments. Press 2 for prescriptions. Press 3 for billing.”

Medicine does not always have a “Press number 4”.

The Receptionist Was Never “Just” A Receptionist

Traditional medical offices have depended on people whose clinical importance was easy to underestimate. An experienced receptionist knew which patient sounded unusually short of breath on the telephone. A nurse knew which message needed to be placed directly in front of the physician rather than left in an electronic inbox. An administrative assistant or clinical manager understood that when another doctor called personally, the issue was probably not just a routine matter--it carried weight and clinical importance.

Such staff members, the critical lever of communication in a physicians’ office, have functioned as the “connective tissue” of a practice.

They recognized voices, knew families, understood which specialist could squeeze in an urgent consultation and which pharmacy needed to be called twice. They often solved problems before those problems ever reached a physician.

Many healthcare systems have stripped away these positions or consolidated them into large call centers in which employees may be hundreds of miles from the hospital or medical office they represent. The result may look efficient on a spreadsheet while functioning poorly in reality.

If a physician spends 10 minutes navigating a telephone tree to reach another physician, that is not efficiency. If a nurse spends part of every shift responding to messages that could have been resolved by trained clerical staff, that is not efficiency. If patients repeatedly call because they cannot reach someone who understands their problem, automation has not eliminated work. It has effectively redistributed work to the most expensive and overextended people in the system.

Patients Experience Healthcare Through Access

Healthcare leaders understandably focus on clinical quality metrics, readmission rates, infection rates and outcomes. But patients often judge a healthcare system through something far more basic: Can I reach somebody when I need help?

A world-class specialist who cannot be reached may be functionally unavailable. A hospital with sophisticated technology but an impenetrable scheduling system can still feel indifferent.

And for older patients, people with disabilities, those with limited digital literacy or patients managing several illnesses simultaneously, forcing healthcare interactions through apps and portals can create a significant barrier--rather than a convenience.

We should also be careful about interpreting lack of patient complaints as indicative of high patient satisfaction scores, or even achieving “magnet status” for a large healthcare system, typically predicated upon specific educational, structural and outcome-driven criteria. The reality is that many patients and physicians “simply give up”---waiting on hold, for a return call or even a text message.

This results in patients and physicians referring care elsewhere. Patients may then appear in an emergency department, leading to ED overcrowding and the adverse clinical outcomes associated with emergency department boarding of admitted patients. Patients may then postpone an appointment, and out of frustration decline to pursue a referral that could be lifesaving or life altering. That failure may never appear as a line item in a quarterly financial report.

Automation Should Support People, Not Replace Them

The answer is not to abandon technology. That would be unrealistic and counterproductive.The goal should be human-centered automation .

Healthcare organizations should automate repetitive work while preserving rapid access to people when situations become complicated. A well-designed system should make it easier—not harder—to reach a human being.

That means restoring adequately staffed front desks, maintaining clear pathways for physician-to-physician communication and giving nurses and office staff enough bandwidth to recognize when an issue needs escalation.

Telephone systems should offer a straightforward route to a person rather than burying it beneath layers of menus. Specialist practices should maintain dedicated professional referral channels so clinicians can communicate directly when necessary. Patient portals should supplement phone access, not become a prerequisite for receiving care.

Most importantly, health systems should stop viewing every human interaction as an expense to be engineered away. In medicine, sometimes the human interaction is the intervention.

A receptionist who calmly helps an anxious patient find the correct appointment matters. A nurse who realizes from a telephone conversation that someone should be evaluated today matters. A physician who can call another physician and discuss a complicated patient for five minutes may accomplish more than dozens of electronically routed messages. These interactions are difficult to measure precisely. That does not make them valueless.

The Same Lesson Applies To AI

As healthcare enters an era of increasingly sophisticated AI, the lesson from our current automation experiment should be clear.

Technology can augment healthcare extraordinarily well. It can also reduce documentation burden, summarize records, detect patterns and automate genuinely repetitive administrative tasks. Used appropriately, AI may even return time to physicians and nurses that technology previously took away.

But healthcare should resist the temptation to use AI primarily as another mechanism for removing humans from patient care. The objective should not be the smallest possible workforce overseeing the largest possible patient population. Instead, it should be the most effective combination of humans and machines.That is an important distinction--and matters immensely.

Medicine Is Still A Relationship

We frequently hear that the healthcare workforce is understaffed. But in many organizations, understaffing has quietly become a standard business model rather than a temporary emergency.

Clinicians are asked to see more patients. Nurses cover more responsibilities. Front-office staff are reduced or centralized. Patients navigate increasingly complex systems largely on their own. Then we wonder why physicians are burned out and patients feel abandoned. Healthcare desperately needs technological innovation. But it also needs something far less glamorous: enough people . People answering telephones, people coordinating referrals, and people helping patients navigate illness and complexities that a phone menu just can’t solve. But most importantly, people giving physicians the time and support necessary to practice medicine rather than endlessly manage electronic queues.

The future of healthcare should certainly contain more sophisticated technology. But progress cannot simply mean automating every interaction that can technically be automated. Sometimes the most advanced thing a healthcare system can do is remarkably old-fashioned-- have a knowledgeable person answer the phone .

Dr. Peter Papadakos , Professor of Anesthesiology and Critical Care, University of Rochester Medical Center, contributed to this article.