America’s medical debt problem is out of control, with collection agencies holding around $140 billion in unpaid medical bills.

Some experts think the amount of medical debt may actually reach $1 trillion and affect nearly 1 in 3 Americans. That’s because not all medical debt appears on credit reports. Moreover, many people pay their medical bills with their credit card and then are forced to have those unpaid bills go to collections.

Beyond the dollars – the human toll is immense. Two-thirds of those with medical debt cut spending on food or clothing and/or have foregone care because of worries about paying their medical bills.

Even scarier, 1 in 7 Americans say they’ve been denied access to medical care because of unpaid bills.

Why is this happening?

The health care system is enormously complex. Hospitals and insurance companies employ teams of people whose full-time job it is to figure out medical costs and medical billing.

Patients who either can’t spend the equivalent of a full-time job advocating for themselves or who don’t have the money to hire their own personal advocate get squeezed.

Furthermore – investment and innovation in health care is all too often focused on insurance companies and hospitals rather than patients (large companies pay more than individual people). This leads to an overfocus on things that can be detrimental to patients (for instance – figuring out better ways for hospitals to collect more money from patients).

Finally – government policy has continually failed to find an answer. And the continued political infighting and gridlock in Washington, DC, means that this failure is likely to continue.

AI as an answer

The recent advancements in AI will change this dire situation drastically – to the point that in just five years – the average American may no longer have to fear debilitating medical debt.

There are two core ways that AI will help:

Automated bill reduction. There are already a myriad of ways to lower a medical bill – from correcting billing errors, appealing insurance denials, finding financial aid opportunities, and just negotiating with hospitals and medical providers to accept a lower payment.

The problem is that the entire system is complex, it’s not always clear what option to use when, and it can be very difficult to navigate the process for each option.

However, with the proper structure and data – AI can be trained to automate this entirely. Imagine a world where you upload a picture of a medical bill, get a customized bill reduction plan, and then sit back as AI gets to work determining what to do and actually advocating and negotiating on your behalf.

Decision-making support. The best way to fight medical debt is to avoid receiving medical bills in the first place. Unfortunately, this requires navigating a Byzantine system of in-network and out-of-network care, covered and non-covered care, pre-authorizations, and medical necessity.

This is hard enough for people whose full-time job it is to figure this out, and next to impossible for the average patient (with their own job, a family, and general life obligations) to do. All too often, this ends up with patients receiving enormous medical bills despite thinking they were covered.

AI will solve this. Imagine an App on your phone that can tell you exactly what needs to be done to ensure that your care is covered, and even automatically work through the process of doing it.

No more waiting on hold with insurance and billing agencies, getting different answers to the same question, and living with anxiety about whether your needed medical treatment is covered. Instead, you can focus on your health and recovery, assured in knowing that you’re fully supported and covered.

What AI can’t do

AI is not a cure-all. It’s not all-knowing, and it doesn’t just magically come into existence. Anyone who has played around with ChatGPT knows that with the amazing things that it can do comes the absolutely weird, odd, and frustrating answers it sometimes gives.

There are also concerns about both privacy and inherent bias in AI algorithms. If you give an AI your sensitive health care data, you don’t want that data being shared to anyone who wants access. In addition, AI is not sentient. Inherent bias in the data and structures that it trains off of will result in biased outputs – which is something we need to be vigilant against.

Finally – AI needs to be properly trained and properly integrated to be useful. AI doesn’t just “know” the right answer – it needs to be trained on the right answer. And then it needs to be built in a way that the average person can easily ask the question to get to the answer.

However, all of this can be overcome with proper thought, focus, and time.

In the near future, AI will help Americans navigate the overwhelming and confusing medical system and provide them with the knowledge they need to make the right financial decisions and avoid the crushing burden of medical debt.

Braden Pan is a health care executive.