For the past several years, much discussion has focused on what AI can do—like faster analytics, automated workflows and predictive models. But there is a more practical question that is ultimately more important: what should AI actually be used for?
In reality, the most effective use cases for AI in healthcare aren’t always the flashiest or most fully automated. Instead, they focus on improving everyday operations—helping teams prioritize outreach, reduce administrative burden and improve the timing of patient engagement.
In other words, AI is proving most valuable when it works quietly behind the scenes to make human interactions more effective.
That insight has major implications for one of healthcare’s most persistent challenges: medication adherence.
The Adherence Problem Hasn’t Changed, But the Tools Have
Medication nonadherence remains one of the most costly and stubborn problems in healthcare. Nearly half of people with chronic conditions do not take their medications as prescribed, contributing to avoidable hospitalizations, suboptimal health outcomes and billions in unnecessary spending.
Despite years of effort, from refill reminders to call-center outreach, many health plans continue to struggle to meaningfully improve adherence rates.
Part of the reason is that adherence has historically been treated as a task rather than a behavior.
Healthcare systems often assume that if patients understand their treatment plan and receive the right reminders, they will follow through. But human behavior rarely works that way.
People skip medications for a wide range of reasons: side effects, financial pressures, competing caregiving responsibilities, transportation barriers or simply the cognitive load of managing multiple conditions.
Sometimes, a patient may have every intention of sticking with their treatment schedule, but their daily routine changes. This context shift makes adherence difficult to maintain because daily habits rely on context clues: such as location, time of day or social influences,
These barriers may be compounded and are rarely static. A patient who is adherent today may fall off track next month because something in their life changes.
That’s why improving adherence requires more than reminders. It requires understanding behavior in relation to real-life circumstances.
Where AI Can Actually Help
This is where AI can make a meaningful difference, not by replacing human engagement, but by making it more efficient and targeted.
In practice, some of the most useful AI applications emerging today are not fully automated systems. They are focused tools that help organizations operate more effectively.
AI can help reduce administrative burden by automating tasks that fall below a clinical license, freeing care teams to focus on patient engagement. It can help prioritize outreach by identifying members who are beginning to disengage before an adherence gap occurs. And it can help to right-size engagement—determining the appropriate timing, frequency and communication channel for outreach so patients are not overwhelmed with notifications they eventually ignore.
But technology alone is not enough.
Behavioral Science Gives AI Context
AI, on its own, is simply optimization at scale. Without context, it can move faster and louder, making it even easier for a stressed or overwhelmed patient to tune out.
This is where behavioral science becomes essential.
Behavioral science helps explain how people make decisions, especially when they are under stress, overwhelmed, facing a new daily routine or managing multiple health concerns. Patients are rarely making decisions in calm, ideal circumstances. They may be juggling work, caregiving responsibilities, financial stress or complex treatment regimens.
Behavioral approaches help simplify choices, frame decisions clearly and anticipate the barriers that patients encounter in real life.
AI can help determine who to engage and when. Behavioral science helps determine how to engage them in ways that actually work.
Technology Cannot Replace Trust.
Even as AI becomes more capable, it cannot replace one of the most important elements of healthcare: trust.
Empathy, relationship-building and clinical judgment remain fundamentally human capabilities that AI cannot replace.
That is particularly true when addressing the real reasons why patients struggle with adherence.
A missed dose is rarely just forgetfulness. It may reflect deeper issues, like a patient who cannot afford both groceries and copays, someone managing depression or a caregiver overwhelmed by competing responsibilities.
Technology can flag these situations. But only human conversations can uncover and address them.
This is why the most successful models combine automation with human engagement.
Designing Systems That Support Real Behavior
Improving adherence also requires rethinking how engagement workflows are designed.
Traditional outreach strategies often place the burden on patients, requiring them to answer calls from unknown numbers, navigate complex refill processes or interpret confusing instructions.
Behavioral science takes the opposite approach: reduce friction and make the desired action easier.
That might mean:
- Simplifying medication instructions
- Offering communication channels patients actually use
- Understanding real-life barriers, such as financial difficulty, food insecurity or a shift in routine
- Escalating to human support when situations become complex
- Personalizing engagement based on individual needs
Ultimately, the goal is simple: make the right action easier than doing nothing.
The Next Phase of AI in Healthcare
As healthcare organizations continue investing in AI, it will be important to focus on applications that strengthen, not replace, human care.
Several emerging areas show particular promise:
- AI-enabled workflow support that automates administrative tasks and frees care teams for patient engagement
- More personalized communication strategies that adapt to patient preferences and behaviors, based on both profile data and context
- Behavioral science integrated into AI-driven engagement models
- Renewed interest in digital therapeutics that combine technology with clinical support between visits
But success will depend on aligning three elements: technology, clinical strategy and the human experience.
A Long-Term Investment in Outcomes
For managed care organizations, particularly those operating in Medicare Advantage, the pressure to deliver measurable, long-term outcomes continues to grow.
Improving medication adherence remains one of the most powerful ways to improve outcomes and reduce the total cost of care. But it requires taking the long view.
Trust is built through consistent, relevant interactions—not through a thousand automated pings.
AI can help healthcare organizations scale engagement more intelligently. However, its greatest value will not come from replacing human interaction. It will come from enabling it, allowing care teams to focus their time and attention where it matters most.
The goal for health plans shouldn’t simply be to use more AI. It should be about aligning technology, clinical strategy, and understanding of how humans make decisions and behave so that, for every member, the path to better health becomes the path of least resistance.
