Part 3 of a series on value-based care
In the first post, I explained why healthcare is splitting into two systems—one that pays for volume, another that pays for value. In the second post, we explored the major players that make up the value based care landscape.
Here's an uncomfortable truth: if you look at the healthcare AI landscape right now, most companies are optimizing a system that's being dismantled.
Ambient clinical documentation. Prior authorization automation. Revenue cycle management optimization. Coding assistants. Scheduling efficiency tools.
These are the darlings of healthcare AI funding. Billions of dollars flowing into making fee-for-service healthcare run more smoothly. And I get it—these are real problems causing real pain for real clinicians. The intentions are good.
But here's what nobody wants to say out loud: you're building a better Blockbuster when Netflix is already winning.
The $4.5 trillion healthcare industry is restructuring around a fundamentally different business model. Over half of Medicare beneficiaries are already in value-based care arrangements. CMS has explicit goals to move the vast majority of healthcare spending to value-based models by 2030. Major health systems are reorganizing their entire operations around population health management.
And most AI companies? They're still building tools to help doctors document visits faster.
Let me explain why this matters—and why the real opportunity is somewhere else entirely.
The Fee for Service (FFS) AI Trap: Incremental Gains on a Broken Foundation
Fee-for-service AI follows a predictable pattern: find a painful administrative task, automate it, save time or money, charge a percentage of savings.
Ambient scribes: Doctor spends a few hours on documentation per day. AI transcribes and structures notes. Doctor gets some time back. Charge $150/month per doctor. To be fair, there is proof that this does lead to lower physician burnout.
Prior auth automation: Health system submits 10,000 prior auths monthly, 20% get denied and require appeals. AI streamlines the process, reduces denials by 30%. Charge based on time saved or appeals avoided. However, this opens the opportunity for payers to build AI systems to combat this and now we have AI fighting AI.
RCM optimization: Hospital has $50M in claims annually, 15% denial rate. AI catches billing errors, improves coding specificity, reduces denials to 10%. Charge 3-5% of revenue recovered.
These are defensible businesses. Some will build successful companies. But let's be honest about what they're actually doing: they're reducing the friction of a fundamentally misaligned system.
Think about what these tools optimize for:
- Faster documentation → enables more patient visits → increases volume
- Better prior auth → gets more procedures approved → increases volume
- Improved RCM → captures more revenue from existing activity → increases volume
Everything points in the same direction: do more things, bill for more things, collect more money for more activity.
This is inherently incremental. You're making existing workflows 10-30% more efficient. That's worth something, but it's not transformational. You haven't changed what healthcare actually does or how it delivers value.
More importantly, you're building on the wrong side of history. Fee-for-service is shrinking as a percentage of total healthcare spending. The more successful value-based care becomes, the less your FFS optimization tools matter.
The Cost Center Problem: Why FFS AI Never Gets Strategic
Here's a dynamic that doesn't get talked about enough: in fee-for-service healthcare, almost every efficiency tool is bought by a cost center, not a revenue center.
Think about who buys FFS efficiency AI:
- Ambient scribes: Purchased by CMOs or physician groups trying to reduce burnout (cost center managing physician satisfaction and retention)
- Prior auth automation: Purchased by utilization management departments trying to reduce administrative overhead (cost center)
- RCM optimization: Purchased by revenue cycle teams trying to improve collections (closer to revenue, but still fundamentally a back-office function)
These are important functions, but they don't drive organizational strategy. They manage costs and operations.
Now think about value-based care infrastructure:
- Population health platforms: Purchased by the C-suite (CEO, CFO, Chief Medical Officer) because they directly impact the organization's largest revenue streams and contracts
- Risk stratification tools: Strategic decisions about where to allocate millions in care management resources
- VBC analytics: Board-level visibility into whether the organization is succeeding or failing at its core business model transition
Having spent years working with VBC organizations, we've seen this dynamic firsthand. In VBC, the technology isn't managing a cost center;it's enabling the entire business model. It's the difference between "nice to have efficiency tool" and "mission-critical infrastructure."
This changes everything about sales cycles, pricing power, competitive moats, and strategic importance. When you're solving a cost center problem, you're always fighting for budget against other optimization tools. When you're enabling a revenue center, you're part of how the organization makes money.
The uncomfortable truth: FFS efficiency AI will always be tactical purchases with limited pricing power. VBC infrastructure can be strategic purchases with enterprise-level pricing because it directly impacts the P&L.
The Uncomfortable Math: Market Size vs. Impact
Let's talk about the actual market opportunity.
The total addressable market for FFS efficiency AI is capped by how much inefficiency exists in the current system. Let's be generous:
-
Clinical documentation: If ambient scribes penetrate 50% of the 1M physicians in the US at $2,000/year, that's $1B annual revenue. The real number will be lower.
-
Prior auth automation: US spends roughly $30B annually on prior auth administrative costs. Even if AI captures 20% of that efficiency, you're looking at a $6B market—and that's split across multiple companies.
-
RCM optimization: The revenue cycle management market is $15-20B. AI might capture 10% over time ($1.5-2B), but this is an intensely competitive space with entrenched vendors.
Add it all up and you're looking at a $10-15B total market for FFS efficiency AI. That's not nothing. But compare that to healthcare's total spending: $4.5 trillion annually.
FFS efficiency AI is fighting over less than 0.5% of healthcare spending.
Now consider the value-based care opportunity: providers are taking direct financial responsibility for $2+ trillion in healthcare spending. Every dollar of unnecessary cost they eliminate, they capture 40-75% of. Every complication they prevent, every hospitalization they avoid, every chronic condition they manage better—that directly impacts their bottom line.
The market isn't capped by administrative efficiency. It's capped by total cost of care.
This is a fundamentally larger game. And this is exactly why we built Pear to focus exclusively on the VBC infrastructure opportunity.
What VBC Organizations Actually Need (And Don't Have)
Here's what changed when we started talking to value-based care organizations instead of traditional health systems.
In fee-for-service, the conversation is always about: "How do we do the same thing faster or cheaper?" The constraint is operational efficiency.
In value-based care, the conversation is: "How do we know which patients will become expensive before they do, and what do we do about it?" The constraint is actionable intelligence at population scale.
After years in the trenches with VBC organizations, we've built Pear to address exactly these questions—the infrastructure that enables healthcare organizations to implement value-based care at scale.
What VBC organizations are trying to do:
-
Proactive risk stratification - "Of our 50,000 attributed lives, which 500 are most likely to have an expensive event in the next 90 days, and why?"
-
Continuous monitoring at scale - "How do we track medication adherence, appointment no-shows, vital signs, and care gaps for tens of thousands of patients simultaneously?"
-
Intervention optimization - "We have 20 care managers. Which patients should they focus on today to maximize impact on outcomes and costs?"
-
Root cause analysis - "Why did our cost per patient increase 8% last quarter? Which conditions, which patients, which providers drove that change?"
-
Predictive care planning - "Based on this patient's trajectory, what's likely to happen in the next 6 months, and what interventions could change that trajectory?"
These aren't workflow automation problems. These are net-new capabilities that don't exist in traditional healthcare because they were never economically relevant.
In fee-for-service, why would you invest in predicting which patients will become expensive? You get paid more when they do become expensive. The incentive is backward.
In value-based care, prediction becomes the entire game. If you can identify a patient heading toward a $50,000 hospitalization and intervene for $5,000 in proactive care, you've just created $45,000 in value—and you capture 40-75% of that in shared savings.
This is where AI becomes genuinely transformational instead of incrementally helpful.
The Infrastructure Gap: What Doesn't Exist
Walk into a traditional health system and ask about their technology stack. You'll hear about Epic or Cerner, maybe Athenahealth. They have EHRs, billing systems, patient portals.
Walk into a value-based care organization and ask what they need. You'll hear about the systems they're building in-house or duct-taping together because nothing exists.
This infrastructure gap is precisely what we're addressing at Pear. Having worked with dozens of VBC organizations, we've seen them all struggle with the same fundamental challenges:
Real-time population health analytics: Not quarterly reports on quality metrics. Live dashboards showing: current run rate vs. benchmark, which patient cohorts are trending expensive, where to intervene today.
Predictive risk models that actually work at scale: Academic research has shown you can predict readmissions, ER visits, and expensive patients with reasonable accuracy. But implementing those models in production, integrating them into clinical workflows, and making them actionable for care teams? That infrastructure doesn't exist as a product you can buy.
Care orchestration platforms: In fee-for-service, you schedule appointments. In value-based care, you're running ongoing programs: transitional care management, chronic disease monitoring, medication adherence tracking, social determinants interventions. You need infrastructure that doesn't exist in traditional EHRs.
Quality measure automation: Medicare Advantage has 40+ Star Rating measures. MSSP has 23 quality metrics. Every commercial contract has different requirements. Manually tracking and closing gaps doesn't scale. You need automated identification, prioritization, and workflow integration.
Financial visibility tools: VBC organizations need to understand their economics at the patient, provider, and condition level. Traditional financial systems can't tell you: "This physician's diabetic patient panel is running 15% over benchmark because of preventable complications." That requires new infrastructure.
Patient engagement at scale: You're not managing episodic visits. You're maintaining ongoing relationships with thousands of patients—medication reminders, wellness check-ins, appointment scheduling, health coaching, remote monitoring. This requires consumer-grade engagement infrastructure that healthcare hasn't historically needed.
These aren't optimizations of existing workflows. These are entirely new categories of infrastructure that become essential when the business model flips from volume to value.
Why This Is Different (And Better) for AI
Here's what makes the VBC opportunity fundamentally more interesting than FFS efficiency AI:
1. The data is richer and more longitudinal
In fee-for-service, you see patients episodically. They come in sick, you treat them, they leave. Your data is a series of disconnected encounters.
In value-based care, you maintain continuous relationships. You have claims history, EHR data, pharmacy records, lab results over time, social determinants information, patient-reported outcomes. This creates a much richer dataset for machine learning.
You're not just predicting "will this note be coded correctly?" You're predicting "will this patient end up in the ER in the next 30 days based on their medication adherence, appointment history, recent lab values, and social circumstances?"
The problem is harder, but the data exists to actually solve it. At Pear, we've built our platform specifically to integrate and make sense of these diverse data streams—something traditional healthcare IT was never designed to do.
2. The feedback loops are real and measurable
FFS efficiency AI has fuzzy ROI. Did the ambient scribe actually save the doctor time, or did they just schedule more patients? Did prior auth automation actually reduce appeals, or did it just shift work from one team to another?
VBC has crystal-clear feedback loops. Did your risk model correctly identify high-risk patients? Did the intervention actually prevent hospitalizations? Did total cost of care decrease while quality metrics improved?
You can measure outcomes in hard dollars: total cost of care, hospitalization rates, ER utilization, readmissions, quality measure performance. This creates a real foundation for machine learning that gets better over time.
3. The value capture is aligned
FFS efficiency AI charges for time saved or process improvements—inherently a small percentage of a small piece of the system.
VBC AI can charge based on outcomes: percentage of shared savings, reduction in total cost of care, improvement in Star Ratings. When an ACO generates an extra $10M in shared savings because your platform helped them manage their population better, a $1-2M software fee seems reasonable.
The value creation is measured in millions or tens of millions, not thousands. This is the pricing model that makes sense for VBC infrastructure—and it's what we've built Pear's business around.
4. The competitive moat is real
FFS efficiency AI is fundamentally feature-based. Your ambient scribe is 5% more accurate than the competitor's. Your RCM tool catches 2% more billing errors.
These are incremental improvements where the next AI model advancement erases your lead.
VBC infrastructure builds moats through:
- Data network effects: The more patients you monitor, the better your predictions become
- Workflow entrenchment: Care orchestration becomes core operations, not a bolt-on tool
- Outcome validation: Proven reduction in total cost of care is sticky and defensible
- Integration complexity: VBC infrastructure must work across claims, EHR, pharmacy, labs, social services—high switching costs
This is infrastructure, not tooling. The differentiation compounds over time instead of commoditizing.
The Counterargument: "But FFS Still Exists"
I can hear the objection: "Fee-for-service isn't going away tomorrow. Even if VBC is the future, there's still a massive existing market for FFS optimization."
This is true. But consider the trajectory:
In 2015: ~20% of healthcare payments were value-based
In 2020: ~35% of healthcare payments were value-based
In 2025: ~50% of healthcare payments are value-based
By 2030: CMS's explicit goal is 80%+ of Medicare in value-based arrangements
Add to this the recent announcement from CMS about the new ACCESS program that explicitly rewards physicians for improving patient outcomes.
Moreover, the organizations succeeding in healthcare aren't those optimizing the old model; they're those building for the new one. When you sell to forward-thinking health systems, they're not asking "how can we document visits faster?" They're asking "how can we manage population health better?"
The innovative buyers are already focused on VBC. If you're building FFS efficiency tools, you're selling to organizations managing decline, not growth. This is why every conversation we have at Pear starts with understanding where an organization is in their VBC journey—because that's where the real strategic investment is happening.
The Real Opportunity: Enabling VBC to Actually Work
Here's what excites me about the VBC infrastructure opportunity: it's not about automating existing healthcare. It's about enabling healthcare that doesn't exist yet to actually work.
Value-based care has been growing for a decade, but it's still operating with duct-tape infrastructure. Organizations are cobbling together:
- Epic's VBC modules (which feel bolted on because they are)
- Third-party analytics vendors (creating disconnected dashboards)
- Custom in-house tools (that don't scale beyond one organization)
- Manual processes that worked for 1,000 patients but break at 10,000
The infrastructure is immature because the category is young. But the trajectory is clear and accelerating.
Having spent years building and refining VBC infrastructure at Pear, we've learned that success requires purpose-built systems—not retrofitted fee-for-service tools. This is why we've focused exclusively on building the infrastructure layer that VBC organizations actually need.
This is where AI becomes genuinely transformational:
Instead of "help doctors document faster," it's "identify which patients are heading toward expensive outcomes and what to do about it."
Instead of "automate prior authorization," it's "predict which interventions will actually change patient trajectories and optimize resource allocation."
Instead of "improve coding accuracy," it's "understand what's driving total cost of care and how to systematically reduce it while improving outcomes."
These aren't incremental improvements. These are new capabilities that change what healthcare can do.
What This Means for Builders
If you're building AI for healthcare, ask yourself:
Are you optimizing the old system or enabling the new one?
Are you making fee-for-service less painful, or are you making value-based care possible at scale?
Are you selling time savings to individual clinicians, or are you selling outcome improvements to organizations managing population health?
Are you building features that get commoditized by the next model release, or are you building infrastructure that compounds in value?
The FFS efficiency market will produce successful companies. But it's inherently capped by the size of the inefficiency and the decline of the underlying business model.
The VBC infrastructure market is capped by total healthcare spending—$4.5 trillion and growing. And more importantly, it's capped by how much better we can make healthcare.
One is about making a broken system slightly less broken. The other is about building the infrastructure for a fundamentally better system.
I know which one I'd rather work on. And after years of working with VBC organizations, I know which one actually transforms healthcare outcomes.
What's Next
In the next post, I'll get specific about what AI in value-based care actually looks like in practice. Not conceptually, but mechanically: what problems are being solved, what data flows where, what workflows are enabled, and what outcomes are being achieved.
Because the opportunity isn't just in understanding that VBC is the future. It's in understanding how to build infrastructure that makes that future actually work.
If you're building in healthcare, the question isn't whether value-based care will dominate—CMS has already decided that. The question is whether you're building for the system that's coming or the one that's being replaced.