Aug 10, 2026
From episodes to journeys: The AI platform for care orchestration
8 min read
Summary
- Building proactive, always-on care journeys means reaching people before they think to ask for help and guiding them through the next steps.
- Agentic AI is already closing care gaps, walking members through Medicaid renewals, booking transportation, and routing patients to the right care.
- The hardest part is the guardrails, escalation paths, and clinical review that make agents safe to put in front of members and patients.
- The organizations making progress are building an infinite care team that scales continuous support without scaling headcount.
Over 40 million people worldwide turn to ChatGPT with a health question every day. Every week, nearly 2 million of those messages are about health insurance: comparing plans, figuring out claims and billing, and understanding coverage. In the U.S., seven in 10 of those conversations happen outside normal clinic hours.¹ When we put those numbers side by side, the gap is hard to miss. People are stressed and looking for help with their health at 11 p.m. on a Saturday, but the organizations accountable for their care aren’t there until the next scheduled appointment or when the call center opens at 8am.
Healthcare has always been transactional. You process a claim, answer a question, book a visit. Each interaction gets resolved in the moment, and then nothing happens until the consumer initiates the next step.
We’ve built portals, call centers, and chatbots to handle those transactions, and most of them do it well. But none of them were built to stay continuously present with millions of people, because that was never something the healthcare industry could pull off.
Let’s picture the alternative: a 24/7 care team that never closes your case, notices what you missed, and reaches out before you even realize you need help. Healthcare has wanted to work this way for years. We just haven’t had a way to deliver it at scale, because the kind of personal attention it requires has only ever come from humans.
Human attention has always been the constraint
If healthcare wanted to scale attention, it had to scale headcount. Every proactive phone call, follow-up on a missed screening, or benefits explanation carries the fixed cost of the human doing the work. It’s the one input in healthcare that has never gotten cheaper at scale, and demand is rapidly outpacing supply. The National Center for Health Workforce Analysis projects a shortage of 141,160 physicians in 2038, and the same trend follows for nurses, care coordinators, and support staff.²
So the chase list has always been longer than the people available to work it. Quality teams know who has open care gaps, but they can’t reach everyone fast enough and the list gets triaged down to whatever the current staffing model can handle.
AI completely changes this equation, because attention no longer has to come from a person. It can hold the context of someone’s medical history, notice what’s missing, start the conversation, and resolve the problem. Suddenly, outreach capacity has nothing to do with headcount.
The cost of that attention is falling fast, too. Querying a model performing at GPT-3.5 level fell from $20.00 per million tokens in November 2022 to $0.07 by October 2024, more than a 280-fold reduction in under two years.³ The attention that used to cost a person’s hourly wage now costs fractions of a penny.
With AI, reaching every member and patient stops being a staffing problem and becomes a design question: What do we want this outreach to do, and how do we guide someone through their care over time?
WHITEPAPER
This generation of AI finally makes the leap possible
See how agentic experiences power always-on engagement, and what it takes to get there.
From episodes to journeys: what an agentic experience looks like
Healthcare has gotten good at resolving episodes. A member calls with a coverage question and gets a quick answer. A patient with a chronic condition visits the ER, gets treated, and goes home with discharge instructions. Both interactions did their job, but neither shaped what happened next.
That customer service rep didn’t know the member had an open care gap, or that she stopped picking up her prescription months ago. The ER staff didn’t realize that the patient never answers unknown phone numbers, and his care coordinator has been trying to reach him since March.
A journey looks different, because it reaches people before they know they need help, ensures next steps actually get done, and remembers that context for the next time. Carrying that context forward is what turns a string of episodes into an actual health journey.
Baptist Health is a great example. Their hello pregnancy app follows expectant and new mothers across their entire maternal journey. Instead of treating pregnancy like a series of disconnected check-ups, the app gives mothers direct access to a nurse and provides guidance tailored to their exact stage of pregnancy. The results speak for themselves: missed prenatal visits fell 11%, well-baby care rose 42%, and postpartum care completion rose 41%.4
Almost everything in healthcare could benefit from this kind of continuous support. Closing a care gap takes time and multiple touchpoints, and resolving a benefits question rarely happens in one chat. That’s where the old transactional model fails people. A member might ask whether her mammogram is covered and get an accurate answer, an in-network provider, and a number to call, but nothing books the appointment for her, checks if she has a ride to the clinic, or follows up to see if she went.
An agentic-powered experience reaches people and stays with them until the job is done. It keeps a running picture of every person, including how they’ve engaged before, what they’ve already been asked, and how they prefer to be contacted. An orchestration layer reads that context in real time and routes the right resources to their needs, working within defined care pathways and routing to a human when it’s necessary.
That’s what makes it possible to stay continuously present and supportive for an entire population.
How agentic AI is already closing care gaps
Consider Maya, a 30 year old woman managing type 2 diabetes while working full-time and raising two kids. Right now, she’s three weeks overdue for a routine A1c test because the reminder came by mail in July and the lab near her office closes at four, before she gets off work.
Under the old model, Maya’s name goes on a chase list. Someone might call her eventually, and if they do, it’ll be from a number she doesn’t recognize in the middle of her workday.
Now, let’s put an agent in her story. Before Maya realizes she’s overdue, the gap gets flagged. The agent pulls her history, finds an in-network lab with Saturday availability near her home, and calls her in the evening when she’s off work. She picks up and it books the appointment. When the agent realizes she doesn’t have a straightforward way to get there, it arranges a ride. One conversation, resolved, with no hold queue or portal login.
Maya is just an illustration, but the capability is real. AI agents across health plans and health systems are already:
- Closing care gaps and arranging transportation to appointments
- Walking families through Medicaid renewal before their coverage lapses
- Guiding members through health risk assessments
- Supporting patients navigating cancer diagnosis and treatment
- Helping families coordinate palliative care
The best part is how the interaction influenced Maya’s outcome. Her A1c results showed she needed a medication adjustment, which means a problem got caught while it was still small. Do that across a population and people live healthier lives. The organization’s numbers change too.
Where most teams stall: building healthcare-grade agents
These kinds of results don’t happen just because an organization bought an AI tool. They happen when agents can be trusted to act, and that trust is rooted in how they’re built.
Back-office AI that gets things right most of the time is genuinely useful, and plenty of healthcare organizations are running it today for claims triage, prior authorization, and coding. An agent talking directly to someone who is sick and anxious has to clear a much higher bar, and it’s the hardest engineering problem in healthcare AI right now. An agent that reaches someone at the wrong time, replies in the wrong language, or hallucinates a coverage detail breaks trust and introduces serious risk.
That’s why teams building AI agents spend most of their time on the foundation. Roughly 84% spend at least half their time on safety infrastructure rather than building the agent itself.⁷ Clinical guardrails and grounding have to tie answers directly to approved sources, expert-authored golden datasets have to act as a master answer key, and evaluation has to run synthetic data against thousands of scenarios to prove an agent is safe and effective to put in front of a real person.
At League, we’ve built Forge by League to give teams that foundation, with guardrails, safety, and compliance built in at the platform level instead of bolted on one agent at a time. The foundation is the hardest part to get right, but it only has to be built once. After that, it’s about deciding which agent to launch first.
FORGE BY LEAGUE
The foundation, already built
Design, build, and run healthcare-grade agents on a foundation with guardrails already in place.
Building a care team that never sleeps
Every health plan and health system already has AI initiatives. Far fewer have agents that are live, trusted, and helping the people they serve. That’s the true gap now.
The organizations bridging that gap are building a new model of healthcare where every person has a care team that never closes the case, knows them deeply, and scales attention without scaling headcount. At League, we call it the infinite care team.
An agent that closes care gaps is a working piece of it, and so is one that walks a member through renewal, or gets a panicked patient an appointment with their PCP instead defaulting to the ER. Each one does real work helping your population while the next one is being built.
The choice facing healthcare leaders is no longer whether to commit to a multi-year platform transformation. The only choice is whether you’re going to start now, with your very first agent.
Here’s the question I’d leave you with: is your organization getting better at managing episodes, or are you building a care team that never sleeps?
Sources
- OpenAI, AI as a Healthcare Ally, January 2026.
- HRSA Health Workforce,Health Workforce Projections, December 2025.
- Stanford HAI, 2025 AI Index Report, April 2025.
- League customer data, Baptist Health, approved for external use.
- PwC, Medical cost trend 2027: Behind the Numbers, June 2026.
- KFF, Medicare Will Spend More Than $13 Billion on the Medicare Advantage Quality Bonus Program in 2026, July 2026.
- Sinch, The AI Production Paradox, May 2026.
WHITE PAPER
The 5 traits of organizations already getting there
Mike Serbinis breaks down what’s separating the healthcare organizations closing care gaps at scale from the ones still waiting for pressure to ease.