A few weeks ago I sat in on a CEO summit, five sessions, a mix of health system CEOs and consumer brand leaders comparing notes on culture, disruption, and AI. Three themes emerged; this is the first.
The Core Tension
Health systems make more money when people are sick. Their mission says the opposite. Nobody on the panel pretended otherwise.
John D'Angelo, CEO of Northwell Health, shared data about what this looks like in practice. Some payers deny claims at rates near 38 percent. Northwell overturns those denials on appeal 99 percent of the time. The claim was right. The system made them fight for it anyway, and that fight costs money, time, and attention that never touches patient care.
Joseph Cacchione, MD, Chairman, CEO & President of Jefferson, put a number on what this does to the bottom line. Jefferson's realization rate on Medicare Advantage contracts runs close to 93 cents on the dollar once denials and administrative burden are factored in. "We lose money on Medicare," he said, describing a system doing exactly what it was built to do.
What Solving It Actually Looks Like
Despite the hype around AI, the fix on display wasn't a new technology, but ownership.
In five words, Amy Perry, CEO of Banner Health, described Banner's own insurance product, covering 1.2 million lives: "We win when they stay well." When Banner is the payer and the provider, keeping someone healthy and keeping the business healthy point in the same direction for the first time.
Northwell has built something similar through a direct-to-employer insurance product, now covering 340,000 lives. The logic is identical. The closer an organization sits to the premium dollar, the more aligned its incentives become. Every layer of sub-capitation between a health system and that premium dollar is a layer where someone else's margin gets built on top of the system's effort.
There's a deadline attached to this, too. The Medicare trust fund is projected to face insolvency by 2032. It's a forcing function already shaping how these systems are restructuring their revenue models today.
Where AI Fits, and Where It Doesn't
AI came up constantly in this conversation, but almost never as the fix for the incentive problem itself. It showed up as the tool for cleaning up the mess the incentive problem creates: denial pattern analysis, administrative automation, faster appeals. AI delivers real value, but as an efficiency tool sitting on top of a structural problem, not a replacement for solving the structural problem.
Efficiency and alignment are not the same fix, and the distinction matters. A system can get very good at processing denials faster and still be losing money on every Medicare patient it treats.
What This Means for PE-Backed Operators
Most operational diagnostics look at cost structure, margin, and process. Fewer look hard enough at whether the revenue model itself rewards the outcome the business claims to deliver.
Three questions to consider before assuming a cost or efficiency fix solves the underlying problem:
Does the revenue model reward the outcome we claim to deliver, or does it reward volume regardless of outcome?
Are we getting faster and cheaper at a process that's misaligned at the root, and calling that progress?
How many layers sit between the work being done and the dollar that pays for it, and whose margin lives in each one?
I've worked with portfolio companies on exactly this kind of gap, where the operational fixes were solid but sitting on top of a revenue model working against itself. If that tension sounds familiar, I'd welcome a conversation about what we found and what changed.
This is the first of three pieces from the same summit. Paper two looks at what this incentive gap looks like from the customer's side.