AI Adoption in Healthcare: The Good, the Bad, and the Ugly

Most of us think of artificial intelligence (AI) in healthcare as something futuristic. We picture humans of the future entering Elysium’s “cure everything” pods, or Robodoc’s heroic robot surgeon performing intricate surgeries to cure formerly incurable diseases.

While there is no “cure everything” pod—not yet, at least—the adoption of AI in healthcare is closer than most of us think. AI is rapidly innovating, and new technological advances are pulling us into a futuristic healthcare world where AI can drive both efficacy and efficiency.

AI has the potential to save healthcare systems millions of dollars, create efficiencies in healthcare administration that streamline processes and systems, and support physicians as they seek better health outcomes for millions of people. It also has the potential to drive up costs and complicate patient care.

When we began planning our first-of-its-kind Certify Blueprint Summit, we knew we wanted Dr. Marc Overhage to speak about the role of AI in healthcare. Blueprint was designed for industry leaders to come together and focus on topics such as AI adoption, burgeoning healthcare administration costs, and health tech innovations. Dr. Overhage possesses a rare combination of clinical expertise, technical vision, and strategic insight.

As Dr. Overhage shared his take on AI-enabled care at Blueprint, we were struck by the nuance of AI adoption in healthcare. AI innovations could be great for our entire system… but they could also have significant repercussions. With this in mind, below you’ll find some of our key takeaways from Dr. Overhage’s keynote and delve a bit deeper into the factors, both positive and negative, that impact AI adoption in healthcare.

The Good

It may come as a surprise to know that AI has been around since the 1950s. Back then, scientists created “machines that could think” by programming them to follow a specific set of rules.

These early AI models could follow a specific set of instructions to do the same thing in the correct sequence, leading to critical technological advancements such as autopilot on 747 jets or the landing sequence of Apache attack helicopters. Our computer keyboards have even been using AI to type letters for decades!

For nearly 75 years, AI evolved slowly, “thinking” through the lens of these programmed sets of instructions. Then, a few years ago, something shifted.

In the last couple of years, Agentic AI has emerged, and along with it, startling potential implications for our healthcare system. Instead of following a rule-based approach, Agentic AI is neural network-based. Essentially, it acts as a giant “brain,” searching a variety of inputs and collecting them into a single, comprehensive dataset.Agentic AI can serve as the eyes, ears, and brains, working alongside a doctor and patient. For example, imagine a pediatrician treating a young child who presents with complaints of general malaise, fussiness, and a low-grade fever. An agentic AI system could “listen in” as the parent described the symptoms and immediately generate a list of potential diagnoses. Then, as the doctor performs various checks, the AI could follow along. AI could “see” the eardrum through the Otoscope and “hear” the heartbeat through the stethoscope. It could collect data throughout the appointment and provide the doctor with real-time feedback on what is happening.

By synthesizing this information, the doctor’s job is easier, the patient is more likely to receive a correct diagnosis, and unnecessary tests or procedures can be avoided, making the appointment and care plan more efficient and effective.

Agentic AI is also helping companies to make considerable strides in streamlining healthcare administration. Many of the systems and processes in place for administrative tasks, such as provider data maintenance, are outdated and cumbersome. Recent innovations in AI-driven tools, such as Certify, are changing that, providing healthcare systems and payers with tech-forward solutions that streamline processes and save millions of dollars.

This future in AI is closer than we think, and we predict that as AI continues to improve, more things will be added to the “good list.”

The Bad

AI is getting better by the day, but there are still some kinks to be worked out. The big one is that “machines that think” are still unable to reason. Dr. Overhage explained it by asking the group to imagine an AI model that had access to all the data necessary to diagnose 100 different diseases. Any patient who walked into the room with the exact symptoms of one of those diseases would get a quick, accurate diagnosis and treatment plan.

But what would happen if someone walked in with disease #101? Or with symptoms from both disease #42 and #74?

A human doctor would be able to reason it out. They would consider the patient’s symptoms and reason that perhaps the patient had two different diseases or an irregular form of a known disease. Human doctors can extrapolate information and make inferences.

AI is fully data-dependent. It can process a large amount of data extremely quickly and manage a substantial amount of information in mere seconds, but it operates on a data-in, data-out basis. If you give it inputs that don’t match the data it has, it won’t be able to reason it out. Even the most advanced AI won’t be able to extrapolate the data to diagnose disease #101.

This inability to reason is on our bad list because human health doesn’t always follow the rules, so to speak. For example, one person may exhibit a list of specific symptoms associated with a given disease, while another person may only exhibit a few of those symptoms. There is hope that AI will get there—some optimistic technologists think that AI could be able to reason within the next five or six years— but for now, the lack of reasoning limits AI’s abilities.

Another “bad” thing about AI is that it’s utterly dependent on having good data to begin with. In the example we gave you above, what would happen if the AI were programmed with the wrong symptoms for disease #38? Patients would get misdiagnosed, treatment plans wouldn’t align, and trust in the technology would be lost.

Certify has addressed the problem of data integrity by using the speed of AI to continuously aggregate our data from thousands of primary sources. This ensures that the data entering our API is accurate and trustworthy, resulting in fewer silos, fewer delays, and lower costs.

The Ugly

There are pros and cons to AI, but a significant, often overlooked truth stands in the way of AI adoption for many healthcare systems and payers: AI is expensive.

While there is a lot of talk about AI “replacing human labor costs” and “creating efficiencies,” at this moment, AI tends to be a cost driver. Despite AI’s rapid innovation, the adoption of AI in healthcare remains slow and tedious.

This has implications for both healthcare systems and insurers. Many administrators and physicians lack access to the best technology to perform their jobs effectively. This makes administration and patient management both expensive and time-consuming.

The cost of AI also has implications for health equity. The cost of AI and uneven distribution across systems means that low-income patients and those from rural areas often lack access to top-quality care.

Conclusion

While we are certain that we will never see a Robodoc performing surgery, the next decade will bring unimaginable innovation in the field of AI. The adoption of highly innovative health-tech infrastructure will lead to lower costs, improved diagnostics, enhanced patient care, and streamlined systems.

Still, if we want to avoid costly mistakes, missed diagnoses, and daunting administration, we need to find ways to democratize access. We need to find the right AI solutions for the correct problems, while leaving human intelligence to do the heavy lifting when it comes to patient care. By doing this, we can move more and more items to the “good” list and eliminate the “bad” and “ugly” parts of AI.

This process can start now. When it comes to provider data management, Certify’s AI-driven, vertically integrated, and centralized platform serves as a single source of truth for provider data. While it may seem like a minor detail in the broader context of the entire healthcare industry, having the correct data in one place has significant implications for healthcare. With Certify, AI handles the “thinking,” allowing humans to focus on reasoning, which saves systems and payers millions of dollars and thousands of hours. Sign up for a live demo today!