2027: The Year Value-Based Care Policy Meets AI

When HLTH opens in Las Vegas on November 15, the expo floor will be wall-to-wall AI: ambient scribes, risk models, copilots, and agents for every corner of care delivery or the revenue cycle. Some of it will be remarkable but only some of those remarkable tools will get adopted.

I'm a technology executive, and I'm also a physician who still sees patients. So I walk the conference hall asking two questions: Will this help clinicians with their patient tomorrow? And will it help the organization carrying the financial risk for that patient over the next decade? Until recently, plenty of organizations could treat the second question as a secondary priority. CMS has now taken that option away and this policy pressure could be pivotal in harnessing AI for care delivery transformation.

The policy signal: more risk, held longer, in more forms

Over the past few years, the CMS Innovation Center (CMMI) has clearly set a course for value-based care to become the normal operating environment for much of Medicare.

Long-term population risk. LEAD (Long-term Enhanced ACO Design) replaces ACO REACH on January 1, 2027, and runs through 2036. That makes it the longest model the Innovation Center has ever tested. It locks each ACO's spending baseline for the full decade instead of rebasing it. It opens the door to small, rural, and independent practices, folds high-needs patients into every ACO, and adds new ways to bring specialists into capitation.

Mandatory episode risk. TEAM went live this January. It makes roughly 750 hospitals in 188 markets financially accountable for five surgical episodes, from the procedure through 30 days after discharge. CJR-X, finalized this summer, goes further. Starting January 1, 2028, every eligible acute care hospital in the country takes on two-sided risk for joint replacement episodes. It's the first time CMS has made an episode model both mandatory and nationwide.

Specialists and chronic care. The Ambulatory Specialty Model, the first mandatory physician payment model, begins in 2027. ACCESS is a ten-year model that pays for technology-enabled chronic care based on outcomes rather than visits. It is CMS betting explicitly that technology, not more appointments, is how chronic disease gets managed.

CMMI’s 2025 strategy further emphasizes prevention, patient choice, competition, and unlocking data in service of CMS's goal of putting every Medicare beneficiary in an accountable care relationship by 2030. The direction for the next five years is clear. Most hospitals and many physician groups either currently hold or will soon hold several kinds of risk at once: a decade-long ACO arrangement, a set of 30-day surgical episodes, Medicare Advantage contracts, and growing specialist accountability. Each has its own attribution rules, benchmarks, quality gates, and penalties. Each one rewards different actions, often for the same patient on the same day.

Why health technology hasn't kept up

Value-based care's technology problem mostly isn't about technology.

For two decades, EHR and the care management platforms have defined our infrastructure. Both are good at recording and billing what already happened. Recently they have connected patient journeys across different sites and have attempted to provide reporting, risk-stratification, and administrative support. However, EHRs were not built to support the care team in determining what should happen next, why, when, and for which patient. Bridging that gap has been slow and expensive. Every new product meant another procurement cycle, another integration project, another login, another analyst to maintain it, and months before anyone knew whether it worked. New companies with good ideas delivering clear value found it difficult to penetrate into the market, and many health systems reasonably decided that the switching or integration cost was not worth it even if it meant sacrificing on optimal technology.

Anyone who has worked in a risk-bearing organization knows the result. Risk scores live in one dashboard, contract terms in a spreadsheet, and care management in a third platform. Clinicians work in the EHR and see none of it at the moment it matters.

The pendulum has swung, hard

Over the course of 2026, this healthtech paradigm has fundamentally shifted. What do I mean? Generative and agentic AI have pushed the cost of building a credible-appearing healthcare tool close to zero. A small team can stand up a risk model, a chatbot, or an "agent" in weeks. That's genuinely exciting when it comes to delivering scalable and high quality care, and it's a big part of why I'm optimistic.

But when anyone can build, the buyer has to do the vetting. Health systems are now flooded with point solutions. Each one solves a thin slice of the problem, needs its own integration work, and adds another stream of alerts for a workforce that is already stretched. Established platforms are running many experiments but despite the potential, ROI has only been established in specific domains of use. Health systems, physician groups, and hospitals are left wondering where to invest and what technology solutions can deliver results. Building is no longer the hard part. Earning trust is.

What clinicians actually need

From the clinical side, the requirements aren't mysterious. They follow from how we actually work.

• Information where we already look. Clinicians take in information in the EHR, on rounds, in huddles, and in the inbox. If a tool needs a new login, it won't get used reliably. Insight has to show up inside the workflow and tech stack we already have.
• Action, not just a score. A risk score usually tells me what I already suspected. What I need to know is which patients I can actually help today, what the next step is, who owns it, and what evidence-based or guideline-driven data drives this recommendation.
• The ability to ask a follow-up question. Clinicians don't work from static reports. We interrogate them. Generative interfaces finally let us ask "Can this patient safely be discharged home?" or "What can I do to reduce the risk of a readmission for this patient?” and get an answer with sources.
• Efficiency, not more work. We face constant pressure on documentation, coding accuracy, and throughput. A tool that adds clicks or needs new staff to run it is fighting a losing battle. The right tools take work off the plate.

What we're building at Radial

At Radial, we think prediction has become a baseline expectation and orchestration is what changes outcomes. We work where value-based care entities meet their physician and hospital partners, because that's where most of the value in these contracts is won or lost. Our platform is developed around 4 pillars:

• Contract intelligence reads each agreement, whether MSSP, Medicare Advantage, or TEAM and CJR-X episodes. It encodes the attribution, benchmarks, quality gates, and penalties so the work lines up with high quality care that the contract actually rewards.
• AI agents for the key levers of value-based care cover admission avoidance, the SNF three-day waiver, hospital and SNF length of stay, readmissions, palliative care, and hospice. They rank patients by risk, actionability, and urgency and show their reasoning. The clinician makes the decision.
• AI-driven care orchestration turns each flag into a next best action with an owner and a deadline, routed into the EHR and care management tools teams already use.
• Next-generation business intelligence lets leaders ask plain-language questions across claims, clinical, and contract data and get cited answers that drive strategic decisions.

The platform connects to the data organizations already have, including claims, ADT feeds, and EHR data, and writes back into their systems. Nothing gets ripped out. It's built on a decade of NIH- and NSF-backed research, and clinicians work alongside data scientists on every agent.

A better on-ramp for technology that works

Providers and patients need a more seamless way to vet and adopt cutting edge technology. That doesn’t mean lowering the bar. Adopting new tools does not have to be an all or nothing proposition. It’s possible to incorporate part of a solution that solves today’s problems if the alternative is the certainty of poor outcomes or care delivery patterns (e.g. readmissions, overutilization of post-acute facility etc.). It’s also possible to use that as a foundation to expand and grow tech-enabled workflows and improve value-based care performance in the future. To accomplish this, we need a better on-ramp, one that gets proven tools to the bedside quickly. In my view, a tool that fits this mold typically has these attributes:

1. Rigor. Grounded in evidence and clinical guidelines, validated against real outcomes, and transparent about how it reaches each recommendation.
2. Built for integration. It fits into the existing tech stack, uses data the organization already has, and doesn't need a new team to run it.
3. Scalability. It works for an independent practice and for a multi-hospital system, and across ACO, episode, and Medicare Advantage contracts. It does not require personnel, though they may be used for specific workflows.
4. Sustainability. It pays for itself in the contract's own measures and delivers a meaningful ROI, so it survives the next budget cycle.

If you're walking the exhibit hall at HLTH, try asking every AI vendor you meet four questions. What evidence is this built on? What will my team do differently on Monday? What do I have to change in my existing tech stack to use it? How will I know, in my contract's own measures, whether it worked? The answers will separate the next decade's infrastructure from excellent ideas that may not get adopted.

Let's talk in Las Vegas

If any of these challenges resonate with you, let’s talk at HLTH. Book a meeting with me. I look forward to an inspiring and energizing meeting!

— Anant Vasudevan, MD, MBA