By Tatiana Fofanova, PhD — CEO, Koda Health

For a long time, we’ve designed care almost like a chess match where the physician — and often the system itself — is the chess player directing the board. The patient moves from treatment to treatment, facility to facility, often reacting to decisions rather than shaping them. The strategy stays focused on fighting disease. The system keeps playing the same game.

You can see the cracks in that model showing up everywhere: in policy debates, in public frustration, and in movements like MAHA. Different agendas, but a shared signal underneath — a growing concern that healthcare is no longer consistently responsive to the people it serves.

Nearly 9 in 10 Americans say end-of-life conversations with their doctor are important. Fewer than 1 in 3 ever have one. Meanwhile, an estimated $200 billion is spent each year on care patients themselves may not have chosen.

Advance care planning, when done well, changes who’s sitting at the board. The patient becomes the chess player. Their values, goals, and definition of quality of life become real clinical inputs — not background information. The provider becomes a guide and partner. Treatment options and care settings become pieces arranged around what matters most to that individual person.

But getting there requires more than a single conversation. Most care for older adults is still episodic. A visit. A transition. A crisis. Then a new set of clinicians with little context for what came before, what the patient wants, or where things are headed.

That’s a problem AI is well positioned to solve: not efficiency at a single touchpoint, but continuity across all of them.

At Koda, we build advance care planning infrastructure. Across partners like Houston Methodist, we’ve seen a 79% reduction in terminal hospital admissions. But the more instructive number is how rarely that outcome traces back to a single conversation or a single intervention. It’s the result of sustained engagement: patients, families, and clinicians staying aligned as health status changes over time. 

What we’ve learned is that ACP has never really been a documentation problem. It’s a follow-through problem. A patient has a conversation. A form gets filed. Six months later, a crisis happens and the family wasn’t aligned, or preferences have changed, or the document exists somewhere no one can find it. The documentation is there (whoopdy-doo?). But the alignment isn’t.

Part of solving that starts with the clinical workflow itself. Our Epic integration lets clinicians send a Koda referral directly from the EMR. Completion data and status flow back into the chart automatically. The clinician doesn’t have to chase an outcome or wonder whether anything happened. And for the patient, that closed loop means something equally important: their plan is part of their medical record, visible to the people who need to act on it, not sitting in a PDF somewhere outside the care team’s line of sight.

This is where longitudinal AI changes the equation. Predictive engagement scoring that identifies who needs outreach before an admission. Behavioral personas that determine not just whether to reach a patient, but when and through which channel. A voice agent that can have a real conversation at 9 PM, because a patient’s readiness to engage doesn’t conform to business hours. AI that follows a patient across time and settings, rather than assisting at a single point and disappearing. That’s where AI moves outcomes, not just operations.

One example of continuity that’s surprised us in practice: AI for decision consistency.

Patients don’t always connect the quality of life they describe to the treatments they select. And a one-time conversation will almost never catch it. Let me give you an example.

A few years ago, I sat with a patient; let’s call her Margaret, 72, recently diagnosed with advanced COPD. She was sharp, opinionated, clear about what she wanted from her life. When I asked what mattered most to her, she didn’t hesitate: independence. Being able to sit on her porch. Watch her grandchildren. She told me, without any ambiguity, that she never wanted to be confined to a bed. That a life tethered to machines wasn’t a life she recognized as her own.

Twenty minutes later, working through her advance directive, she selected ventilation in all circumstances.

She didn’t catch the contradiction. Why would she? Margaret wasn’t a clinician. She’d never had to think about what ventilation actually looks like — the tube, the sedation, the ICU bed she’d never leave. To her, “ventilation” sounded like help breathing. It sounded like a bridge back to her porch.

This is the gap that doesn’t show up in quality metrics. Margaret’s advance directive was technically complete. Box checked. But it didn’t reflect what Margaret wanted — it reflected what Margaret understood, which through no fault of her own, was incomplete.

Longitudinal engagement surfaces exactly this. AI helps us catch the inconsistency, prompt the right follow-up, and use targeted education to close the gap between what someone values and what they’ve actually elected. Most people don’t live in the world of medicine. They shouldn’t have to. But they deserve something that bridges that gap before a crisis makes the question urgent.

Continuity isn’t just about staying in touch. It’s about the quality of understanding that builds when you do stay in touch.

That leads to something bigger than workflow. There’s a real and growing frustration in this country with healthcare institutions and a growing sense that patients don’t have sufficient voice in decisions that directly affect their care. Misaligned care is, at its core, what happens when the system stops listening. That frustration is showing up in policy conversations, in reform movements, and in what we hear from patients themselves.

Advance care planning, done well, is one of the most direct responses to that problem. It returns voice to the patient. It treats what matters to them as a clinical variable, not paperwork, not a checkbox, but information that should shape care in real time and keep shaping it as things change.

AI gives us the infrastructure to do that at scale. Not by replacing the human conversation, but by ensuring it happens and keeps happening. A patient’s values shouldn’t live in a file. They should follow the patient across every point of care, evolving as their health and priorities change.

That’s the continuity opportunity.