We did report cards the old-fashioned way. The envelope showed up, my mother opened it at the kitchen table, and there was a period of silence that felt roughly nine hours long before anybody said anything.
I have been thinking about that silence a lot lately, because I wrote a book called RCM 2030 full of confident sentences about what revenue cycle would look like by the end of this decade, and 2026 has started grading it.
I got some things right, and I got some wrong. And two of my predictions were wrong in ways I find genuinely interesting, which is the only kind of wrong worth writing about.
So here's the envelope. Let's open it together.
The short version
What I got right: Epic swallowing the revenue cycle stack, payers weaponizing AI faster than providers, arbitration middlemen who profit from your errors, cybersecurity as a cash flow problem rather than an IT problem, and the coming collapse of the entry-level analyst role.
What I got wrong: I said Epic wouldn't build meaningful patient financial tools before 2030. They shipped several this month. And I built a whole vendor-survival forecast without once considering that Epic might get into the staffing business.
What I flat out missed: Ground ambulance is not covered by the No Surprises Act. I have written seven books about revenue cycle and I learned this in August 2026, which is humbling in a way I'll explain in a moment.
What I got right
Epic would absorb the bolt-on vendors
I wrote that by 2030, hospitals would stop paying outside vendors for basic eligibility and prior authorization, that Epic and Oracle would make payer APIs standard inside the revenue cycle stack, and that clearinghouses would have to pivot into analytics or disappear. I also put "standalone payment portals" at the top of a list called Categories Likely to Consolidate or Exit by 2030.
At its August Users Group Meeting, Epic announced Savvy, its own native patient payment platform. Also autonomous coding, Payer Policy Insights, automated underpayment recovery for 2027, and an integrated eFax platform, which I mention only because somebody at Epic had to sit in a meeting in 2026 and talk seriously about fax.
Grade: A. The direction was right and the timing was close.
Payers would use AI to deny faster than providers could bill
This one I wrote with some heat. Payers were already ahead, they were auto-denying at volume, and hospitals were going to have to meet fire with fire.
The first full year of public denial data is now out. Insurers denied somewhere between 12% and 18% of prior authorization requests across Medicare Advantage, Medicaid managed care, and the ACA marketplace. When patients appealed, 67% of Medicare Advantage denials were overturned. Centene overturned 93% of its appealed MA denials.
93% overturn rate. Yikes. That’s not a claims review process. It’s a toll booth with a very slow attendant.
Grade: A.
The middleman who makes money on your mistakes
In the Operations Companion Guide, I described a category of vendor that survives on dysfunction: outsourced denial recovery firms paid contingency fees who "recover some cash, but they make money on your errors."
Georgetown researchers published an analysis in Health Affairs this month putting the total cost of the No Surprises Act's arbitration process at $22.4 billion through the end of 2025, and they call that number conservative. Three organizations account for roughly three quarters of all resolved disputes. One of them, HaloMD, does not treat patients at all. It exists to file.
The federal government projected about 22,000 arbitration cases a year when it built this thing. 2025 produced 2.6 million.
Grade: A, and I wish I'd been wrong.
Cybersecurity would become a margin issue, not an IT issue
I wrote that attackers would eventually use AI to exploit vulnerabilities the minute patches rolled out, that every vendor you add is another doorway, and that a 48-hour outage in 2030 would hurt the way a two-week outage hurts today.
In August, Epic told its own customers that the 30-day patching standard "is no longer good enough," and disclosed that it has been using a restricted-access AI model to hunt for flaws its own developers missed in a several-hundred-million-line codebase. That same week, more than a hundred technology and security companies signed a public letter naming hospitals specifically as critical infrastructure facing a surge in AI-enabled attacks.
Grade: A minus, only because I framed it around ransomware and the actual 2026 shape is broader than that.
The entry-level analyst was in trouble
I built a whole Skills-by-Role matrix arguing that a billing specialist in 2030 would spend half the day reviewing machine-generated output, and that coders and CDI staff would move from validating codes to overseeing algorithms.
Epic unveiled a tool this month designed to do junior analyst build work in minutes. The CIO of Advocate Health responded publicly with the exact question I'd been asking: how do we grow the senior analyst of the future when the junior work is gone?
Grade: A. Cold comfort, but an A.
Nobody would be able to prove the AI worked
"No technology strategy survives a talent gap" was my line, along with a dashboard note that if the technology isn't improving yield, it's shelfware.
Deloitte surveyed 64 healthcare CFOs and finance leaders this year. Forty-four percent qualified as "AI scalers." Of those, only 18% had mature financial attribution, meaning they could actually tie AI to revenue or cost with defined baselines and clear ownership. Organizations earlier in their AI journey scored better, at 31%.
Sigh. The further along you are, the less able you are to prove it worked.
Grade: A.
What I got wrong
Now the part where my mother would have gone quiet.
I said Epic wouldn't get consumer-centric before 2030
Direct quote from my own book: "If you're a CFO banking on your investment with Epic to solve for consumer-centricity, you may be waiting beyond 2030."
I based that on thirteen years of building products meant to live in Epic's marketplace, and on a pattern I'd watched hold for a decade. Epic builds for where the money flows, and the money flows from payers and the government, not from patients.
Then this month, they shipped a native patient payment platform, patient-facing scheduling through their AI assistant, and Medicaid application assistance, which is functionally the thing I have been shouting about for years: screen the patient for coverage and assistance before the balance turns into bad debt.
Here's what's interesting, though. My reasoning was correct and my clock was wrong. I said the forcing functions would be regulation, competition, and the fact that the underlying technology already existed. All three did the work. They just did it faster than I gave them credit for.
And the deeper critique still stands. Nothing Epic shipped thinks in terms of lifetime value or household relationships. It's still transaction-level tooling, built one encounter at a time.
Grade: C. Right about why, wrong about when, and "wrong about when" is not a small thing when CFOs are building five-year capital plans off your book.
I never considered that Epic would become a staffing company
I wrote a whole section forecasting which RCM vendors survive the next five years. I named the big platforms, predicted AI-native startups, floated Amazon and CVS as potential entrants, and covered whether EHR vendors would build RCM in-house or open app-store ecosystems.
What I did not predict, not even a little, was Epic Rangers: Epic-hired, Epic-trained, Epic-payrolled employees embedded long-term inside a customer's own IT department.
That's not a software play. That's a services play, and it lands in a consulting market that private equity has already been rolling up and offshoring. Every health IT consulting firm in the country now has a competitor with 340,000 job applicants a year and a hiring bar they've spent four decades enforcing.
Grade: D. I was looking at the software market, and the threat came from the services side.
What I flat missed
Ground ambulance is not covered by the No Surprises Act.
Air ambulance is. Congress carved out ground and handed it to an advisory committee.
I learned this in August 2026. Seven books in.
The most involuntary purchase in American healthcare, the one you make by dialing three numbers during the worst ten minutes of your life, in a vehicle you cannot choose, staffed by people whose network status you will never know, is the one place the surprise billing law does not reach.
I'm not going to pretend that's a prediction I got wrong. It's a fact I didn't know, and the reason I'm putting it in a public report card is that I'd bet real money most revenue cycle leaders don't know it either. If you're going to write about how a law protects patients, you should probably know where it stops.
What I'm still betting on
Since I'm handing out grades, here's what's still ungraded, and where I'd put my money today.
Denial rates below 5% by 2030. Still my call. Still ambitious. The prevention technology is arriving faster than the operational discipline to use it, which is the actual bottleneck.
Financial AI disclosure becomes mandatory between 2028 and 2030. New prediction, and the one I feel most strongly about. Pew found that 72% of Americans want to be told when AI is used in their care, and 46% have no idea whether it already has been. Clinical AI disclosure standards are coming. Financial AI gets swept in, either by accident or by a state attorney general who decides autonomous coding materially affects patients.
Congress reopens the No Surprises Act. At $22.4 billion and 2.6 million disputes a year against a 22,000-case projection, this stops being a policy question and becomes an arithmetic one.
The employer becomes the loudest voice in the room. Sixty-four percent of employers surveyed by the Business Group on Health this year reported a measurable cost impact from providers' AI-driven revenue optimization. Health systems can't prove their AI saves them money. Employers can already prove it's costing them. That asymmetry is going to matter enormously, and I did not see it coming in the book.
Frequently asked questions
What is RCM 2030? RCM 2030: Strategy and Survival for Revenue Cycle Leaders is a book I wrote for hospital CFOs and revenue cycle executives, forecasting how AI, interoperability mandates, and federal policy will reshape hospital billing by 2030. It's part of a seven-book series that includes companion guides for operations, policy and regulation, and workforce modernization.
Will AI replace revenue cycle jobs by 2030? Not wholesale, but the shape of the work changes hard. Repetitive claim handling, basic eligibility checks, and junior build work are the most exposed. Roles requiring algorithm oversight, payer logic literacy, exception handling, and direct patient financial communication are the ones that hold value. The genuine risk is not mass unemployment; it's that organizations automate the entry-level work and then discover in five years that they have no pipeline to senior expertise.
Is Epic taking over revenue cycle management? Epic has moved aggressively into territory that used to belong to third-party vendors, including patient payments, autonomous coding, payer policy analysis, underpayment recovery, ERP, and long-term IT staffing. Whether that's takeover or consolidation depends on where you sit. If you sell a point solution in a category Epic just entered, it feels like one thing. If you're a CFO paying for eleven bolt-on vendors, it feels like another.
How much has the No Surprises Act arbitration process cost? Georgetown researchers put total costs at $22.4 billion through the end of 2025, including $15.6 billion in payments above in-network rates, $4.2 billion in administrative costs, and $2.7 billion in fees to arbitration entities. The researchers describe their estimate as conservative.
Does the No Surprises Act cover ambulance bills? Air ambulance, yes. Ground ambulance, no. Congress excluded ground ambulance services and directed federal agencies to convene an advisory committee on the issue. A number of states have enacted their own ground ambulance balance billing protections, but those generally don't apply to self-funded employer plans.
The thing about report cards
The point of grading yourself in public isn't self-flagellation, and it isn't a victory lap either. It's that a forecast you never check is just a vibe with footnotes.
I got the direction right more often than I expected and the timing wrong in the two places it mattered most, both in the same direction: things are arriving faster than I thought they would. If there's a lesson in the envelope, that's it. Every time I've been wrong this year, I've been wrong by being too conservative about the clock.
Which means if you're reading my books and thinking "sure, but we've got time," you and I are making the same mistake.
I write about this every Sunday in RCM 2030 Weekly, where hospital CFOs and revenue cycle leaders get the week's policy and AI news filtered through what it actually does to your A/R. Come sit at the kitchen table with me.
