As 2025 closes, I’ve felt a growing urge to write more. I spend my days building technology around value-based primary care, and I keep seeing the same pattern: the biggest constraint isn’t ambition, or even ideas. It’s the missing infrastructure that turns fragmented data into coordinated action inside real workflows.

In 2026, I’m going to write about that layer.

I am a part of engineering at Aledade, working alongside independent primary care practices that are taking on value-based accountability. That vantage point shapes how I see the world. You can debate which model should win, which policies will stick, and which vendors will capture share. But on the ground, the bottleneck is surprisingly consistent: coordination is still expensive, brittle, and labor-intensive.

If you zoom out, 2025 looked like a year of convergence. Three forces that used to move independently started to line up.

1) Value-based care stopped feeling optional

The story of value-based care is often told as an aspiration. In 2025, it increasingly read like an economic inevitability.

When providers are accountable for outcomes and total cost of care, the math changes. Preventable admissions and avoidable ED visits are not just clinical failures. They are margin destroyers. Prevention and chronic care management stop being “nice to have” and become the only coherent strategy.

We have seen this direction for years, but the footprint keeps expanding. CMS has publicly highlighted continued savings and quality performance in Medicare accountable care programs, including the Medicare Shared Savings Program’s net savings in recent performance years. (Centers for Medicare & Medicaid Services) And across the industry, estimates suggest tens of millions of lives are now in some form of accountable care arrangement across payers. (UnitedHealth Group)

The more this footprint grows, the more the same constraint shows up: you cannot run proactive, population-level care on workflows that still depend on delayed data, manual reconciliation, and heroic effort.

2) Interoperability became less theoretical and more operational

Interoperability in healthcare has been a long series of “almost” moments. Standards existed. Networks existed. Policy existed. But getting the right record to the right workflow at the right time was still inconsistent.

What felt different heading into 2026 is that the substrate is starting to look more like an actual exchange layer, not just a set of standards and good intentions.

TEFCA is a useful marker here. The Recognized Coordinating Entity (RCE) has reported rapid growth in the number of organizations live on TEFCA and the volume of documents exchanged since go-live in late 2023. (ASTP TEFCA RCE) And the number of designated QHINs has continued to climb through 2025. (Healthcare Dive)

At the same time, the era of “closed gardens by default” is getting harder to defend. HHS has signaled increased enforcement posture around information blocking. (HHS)

None of this means “data is solved.” Anyone building real clinical workflows knows it’s still messy. But it does mean the question is changing from “can data move?” to “can we operationalize it safely, consistently, and cheaply enough to matter?”

3) AI moved closer to workflow, and closer to the line of accountability

2025 was full of AI activity, but the most important shift was not model capability. It was placement.

AI is steadily moving from standalone experiences into embedded workflow, and that matters because embedded AI inherits the responsibilities of the workflow it touches.

In documentation, that looks like ambient and assistive tools that reduce clerical burden and help clinicians stay present. In operations, it increasingly looks like automation of repetitive, rules-heavy processes like prior authorization and utilization management, where the stakes are real and delays can cause harm.

You can see this shift reflected in policy experiments and pilots that explicitly blend technology with administrative gatekeeping and payment incentives heading into 2026. (MarketWatch)

This is where the world view and my day-to-day perspective collide:

  • AI is getting easier to access. But healthcare is not a domain where access equals readiness.
  • In 2026, the winning teams will not be the ones with the best model. They will be the ones that can make trust rational.

The three beliefs I updated in 2025

Belief 1: Interoperability is necessary, but not sufficient

Interoperability is a prerequisite. It is not an operating model.

Even when data is technically available, you still have to solve: identity, consent, normalization, provenance, routing, and accountability. “FHIR exists” is not the same thing as “the right context shows up for the right person at the right time.”

2025 reinforced for me that healthcare does not have an innovation problem. It has a translation problem.

Belief 2: Primary care is the coordination layer, but it has not been given coordination infrastructure

We keep asking primary care to manage the longitudinal journey: prevention, chronic disease, behavioral health, referrals, transitions, end-of-life planning. That only works if primary care has leverage.

Independent practices rarely have the technical staff or spare operational bandwidth to stitch together portals, labs, claims, and hospital feeds, then prove outcomes, then redo it every time a new tool arrives.

What they need is a platform, not another dashboard.

A platform that makes coordination cheaper than fragmentation. A platform that turns “we should” into “we can” without burning out the people doing the work.

Belief 3: Governance cannot be a committee. It has to be a product

Healthcare governance that only exists as a quarterly review is governance that fails daily.

The only scalable path is to treat governance as product. Encode the rules. Automate the guardrails. Make the compliant path the path of least resistance.

This is the heart of “earned AI.”

If you cannot explain, audit, monitor, and safely degrade an AI-driven recommendation inside the workflow, then you have not built a clinical capability. You have built a liability.

What I think 2026 will reveal

Here’s my world view bet for 2026, shaped by what I saw in 2025.

1) “AI everywhere” will meet the reality of accountability

AI will continue to spread through EHRs and operational tooling. But the center of gravity will move from novelty to evidence.

Where does it demonstrably reduce burden? Where does it demonstrably improve quality or outcomes? Where does it introduce risk, bias, or delays? Who is responsible when it fails?

The loudest conversations will not be about prompt quality. They will be about governance, measurement, and rollback.

2) The administrative layer will be an early battleground

Prior auth, claims workflows, and utilization management are tempting targets for automation because they are repetitive and rules-driven. They are also a place where automation can create invisible harm if it is wrong or if it shifts friction onto clinicians and patients.

The most important question is not “can AI do it?” It is “can AI do it with transparency, appeal paths, and safety guarantees that preserve trust?” (MarketWatch)

3) Rural and independent care settings will matter more than ever

If you care about outcomes and equity, the long tail is the point. Rural and independent practices are where a huge amount of longitudinal care happens, and they are often the least resourced to adopt bespoke tooling.

Recent federal and state-level initiatives heading into 2026 underscore that rural capacity and transformation are front-of-mind, even if the politics and implementation details are contested. (Reuters)

If healthcare is going to improve at scale, infrastructure has to work for the long tail, not just flagship systems.

4) The next differentiation will be utility layers, not point solutions

The last decade produced many valuable point solutions. The next decade will reward shared utility layers: identity, consent, data normalization, auditability, workflow routing, evaluation, monitoring.

When those layers exist, innovation compounds. When they do not, innovation resets to zero every time you change an integration, a policy, or a workflow.

What I’m going to write in 2026

This post is the trailer. Here are some of the essays I plan to publish next:

  • Primary Care Needs a Platform Why the coordination layer of healthcare needs shared utility layers, not more one-off tools.

  • Governance as Product How to make the safe path the easy path, with policy-as-code, auditability, and default guardrails.

  • AI Is Earned, Not Deployed A practical view of what “trustworthy clinical AI” actually requires: clean data flow, feedback loops, safe degradation, and human override.

  • Workflows Beat Dashboards Why the unit of change is not an analytics view, but a routed task inside a real team’s day.

  • The Long Tail Problem How you ship modern capabilities across thousands of independent practices without custom integration and bespoke compliance every time.

If you are building in this space, I would love the pushback. What is the most important thing that needs to become invisible? What part of the workflow feels most broken? Where do you think the industry is overconfident?

I’m going to write toward the messy middle, because that is where the work actually is.

Disclosure: I work at Aledade. Views are my own.