Dear Sir or Madam,
My name is Jonathan Gluck. I’m a cancer patient. I have a type of blood cancer called multiple myeloma that is treatable but, in all but a small percentage of cases, incurable. I’ve been living with my illness for more than 22 years. During that time I’ve undergone dozens of rounds of radiation therapy, chemotherapy, and immunotherapy as well as a difficult and potentially fatal treatment called CAR T-cell therapy. I’ve gone in and out of remission more times than I care to recall. And I’ve spent thousands of hours in doctors’ offices and hospitals, filing thousands of health insurance claims worth millions of dollars.
I am writing with an urgent request. Please don’t deploy AI customer service agents. And if you’re already doing so, stop.
The current landscape
For some time now, leaders in industries from e-commerce and financial services to travel and hospitality have been employing AI agents to interact with their customers to increase efficiency, cut costs, and boost profits. Many large health insurers are using AI agents, as well, for settling claims disputes, ruling on prior authorization requests, and managing other functions.
I won’t hide my bias. I hate AI agents. At best, I find interactions with them to be cold and soulless. At worst—when something goes wrong, say, or the agent doesn’t understand your request—they are infuriating. When you’re trying to buy an airline ticket or return a pair of flip-flops, that’s bad enough. But when you’re dealing with a health issue, especially a serious one, it’s not just frustrating. It’s reprehensible. If you take humans out of the loop, you’ll be removing patients’ last lifeline to caring, decent treatment.
Firsthand experience
My run-ins with health insurance difficulties date back to 2003, when I was first diagnosed with cancer. At the time, my doctor ordered a PET scan to confirm my diagnosis and determine if my disease had spread.
X-rays, MRIs, and CT scans can “see” anomalous spots in bones and soft tissue, but they can’t detect metabolic activity. A PET scan can. In the procedure, a radioactive tracer is injected into your bloodstream. After waiting for the solution to circulate throughout your body, you lie down on a machine similar to a CT scanner and slowly pass through it for 20 minutes or so. Because cancer cells “take up” the radioactive tracer faster than other cells do, malignancies light up.
Never mind that a PET scan is considered among the most valuable tests for multiple myeloma and that detecting the disease early might add years to my life. Oxford, my insurance company at the time, refused to pay for it. In their twisted logic, I had no evidence of multiple myeloma from the tests I’d done so far, so I didn’t have it. No further testing was needed. It was the first of what would turn out to be many insurance battles I would fight over so-called pre-authorization.
A thorny process
Pre-authorization, as you well know, is an attempt on behalf of companies like yours to limit unnecessary tests, medications, treatments, and procedures. Before you will agree to pay for certain items, especially high-cost or high-risk ones, your in-house medical experts review them to determine whether they consider them necessary. If so, the procedure is approved. If not, not.
In theory, the system exists to protect patients from undergoing unnecessary tests, treatments, or procedures and to keep health care costs in check. In practice, it is also used as a way for your companies to enhance their profits. It also creates red tape and can slow down, if not block, patients’ access to healthcare.
Beginning with that PET scan decision, and continuing to this day, I have had to appeal scores of pre-authorization denials.
As anyone who has gone through the process knows, it can be Kafkaesque. Essentially, your doctor’s office needs to convince the insurance company that the test or treatment in questions is, in fact, necessary. That involves the doctor’s office sending a letter to the insurance company making that case, along with the appropriate medical records.
That process can take days or weeks under the best of circumstances, and far too often something goes wrong. The doctor forgets to ask the pre-auth coordinator to make the request. The pre-auth coordinator is slow to pursue the matter or forgets about it altogether. The doctor’s argument isn’t convincing. The insurance company is slow to respond. All of those things and more have happened to me. And that was with humans in the relevant jobs.
Ultimately the burden falls on the patient to resolve any difficulties. The time that takes and the stress it brings can be considerable. You often have to call the doctor’s office and the insurance company repeatedly, and spend countless hours on hold. Frequently you’re directed to the wrong person or department. “Unfortunately,” the person you reach isn’t sure why the procedure was denied. All of this happens when you are sick or otherwise in need of medical care and at your most vulnerable.
What works
I have personally spent hundreds of hours fighting pre-authorization battles and have tried every trick I can think of—calm reasoning, obsequious kindness, screaming—to resolve them.
What I have discovered is, the person on the other end of the line has to care. Sometimes they do, sometimes they don’t. It turns out the secret to navigating the pre-authorization maze, if not the American health insurance system in general, is to cross your fingers and hope you get a caring human on the other end of the line.
If you take humans out of the equation, even that slim opportunity will be lost. And the cumulative amount of additional unnecessary time, effort, and emotional distress that will follow, across billions of customer interactions, makes me sick to imagine.
A recent case
It’s not just pre-authorization that’s the problem. The same dynamics apply to claims disputes.
In August, I filed a claim for $1,000 for four visits to a therapist I made in July. (I see her once a week to help manage the anxiety and depression my illness has precipitated.) My therapist happens to be out of network with my current insurer, and she offers only telehealth visits. Some health insurance plans don’t reimburse under those circumstances and for a time, mine didn’t either. But when I told my HR representative about the problem, she raised the issue with my company’s management team and they were kind enough to add that coverage to our policy. As of January 1 of this year, I was supposed to be covered. And in fact, I was. My claims for January through June went through as they should.
When I received the Explanation of Benefits for the July claims, however, I saw they had been denied. The stated reason was that my policy doesn’t cover out-of-network telehealth visits. Only it does—and my insurer had already reimbursed me six times, in January through June, for just such appointments.
After multiple discussions with claims representatives, I got nowhere. When one of them hung up on me, I went back to my HR representative. I have since been told the claim has been reprocessed and I should receive payment within 30 days of the reprocessing date. That would put my reimbursement date some 90 days later than the date I paid my July therapy bill. I don’t get 90 days to pay my insurance premiums. Why should you get 90 days to reimburse me for the services I use—and pay you tens of thousands of dollars a year for?
Again, all of that happened with humans in the equation. The idea that AI agents will do anything but make the claims dispute process even slower and more upsetting than it already is, no matter how “smart” AI agents eventually become (and they’re not currently very smart), is laughably naive.
It’s no secret that millions of Americans have a dim view of health insurers. As medical consumers, we are fed up with cold, impersonal treatment by your industry. Deploying AI agents will only make it colder and more impersonal.
The way forward
I wrote that the secret to navigating the U.S. health insurance system is to hope you get a caring human on the other end of the line. After one particularly long and drawn-out ordeal over a mishandled claim a number of years ago, I threw myself at the mercy of maybe the eight or ninth claims representative I had spoken with. I told her I normally don’t like to play the cancer card because I feel it’s manipulative. But I also told her I was in the middle of a round of treatment and I was physically and emotionally exhausted. I asked her if she would please, please, take a personal interest in my case, and resolve it herself, no matter what that took, without requiring any more phone calls or emails from me.
Incredibly enough, she said she understood. She told me that she had lost her mother to breast cancer earlier that year and that she was sorry for the hassle I had been through. She connected me to a supervisor, explained to her what had happened to me to date, without me having to repeat it again, and stayed on the line until the matter was resolved. I got the reimbursement check by FedEx two days later.
AI agents don’t have mothers with cancer. They don’t have empathy. Whatever else they can do, they cannot care. (Cheap automated sycophancy doesn’t count.)
Health insurance consumers need—and deserve—kind, compassionate, and timely service from you and your companies. Take it from someone who knows. The most powerful technology for that is a simple, decent human.
Jonathan Gluck is a Fast Company senior editor and the author of An Exercise in Uncertainty (Harmony Books), a memoir chronicling the more than 20 years he’s been living with cancer. Parts of this essay are adapted from that book.
