Why Does AI Keep Creating More Work for My Revenue Cycle Team?
I have been telling finance leaders the same thing for years, and this month the evidence piled up to prove it. Automation does not shrink your team. It changes what your team has to be good at. Miss that, and you will cut the wrong people at exactly the wrong time.
The goal was never fewer humans. It was different humans, doing higher-value work, with enough headcount to actually do it. Most hospitals budget for the first half of that sentence and pretend the second half doesn't exist.
What did experts actually say about AI and jobs this month?
Jeff Bezos told CNBC the employment doomsayers have it backwards, that AI is pushing us toward a labor shortage rather than a glut, and that it will elevate people instead of replacing them. Hand someone a bulldozer, he said, and they don't dig less; they take on a bigger job.
Dan Shipper, who runs an AI company called Every, said it better, because he brought a number. His company automated everything it could with AI agents. Then it grew from four employees to more than thirty. Not despite the automation. Because of it.
His breakdown is the clearest I've read. Humans set the frame at the start: what are we doing, what counts as good. The AI collapses the middle: it drafts, searches, summarizes, compares. Humans judge and extend at the end: is this right, where does it belong, what happens next. The machine takes the middle. The valuable, defensible work moves to the two ends.
Why does automating a task create more work instead of less?
Picture a cardiology patient whose scan gets read by an algorithm that catches a risk a human would have missed. Good. Now what? Somebody orders the echo. Somebody reads it. Somebody gets the patient back in. A cardiologist at Penn Medicine said it plainly this month: there aren't enough humans, and leaning on humans-in-the-loop as your safety net stops working when the loop keeps getting longer.
At Jefferson Health, a team built AI to comb a year of MRIs for missed neurological conditions and bring those patients back. It worked. It came back with more than the team could handle. Stanford now runs every rollout through one question: before we launch this, how many more specialists will we need to act on what it finds? Do that math first. Then deploy.
This is the part the vendor deck skips. AI is great at finding work. It is not good at the part that needs a human to decide and act. So every tool that finds more, flags more, and generates more creates a wave of downstream human work that has to land on somebody.
What is the hidden cost of AI that nobody put in the budget?
A study out of Penn Nursing this month named it. AI tools can raise cost and cognitive burden instead of lowering them. They need training, maintenance, workflow redesign, oversight, security, and constant evaluation. The researchers flagged "algorithmic drift," where a model that worked at launch slowly stops working as your patients and workflows change, which means somebody has to watch it forever.
None of that lives in the line item marked "AI savings." In 2020 the cost of a tool was the license. In 2025 it's the license plus the people who train it, validate it, and catch it when it drifts. By 2030 the shops still standing are the ones who budgeted for the second number from the start. The one line I'd staple to every budget: no technology strategy survives a talent gap.
What happens to my revenue cycle if I automate without staffing for it?
The same thing that happens on the clinical side. Your AI clean-claim checker catches more issues; somebody has to work them. Your denial-prediction model flags more at-risk claims; somebody has to act before submission. Your coding AI produces more output; somebody has to be the quality lab that catches the odd answer before it turns into a payer dispute.
In the Workforce Modernization Guide I mapped where these roles go. The coder becomes an AI quality analyst. The collector becomes a patient relationship agent. The analyst becomes an automation architect. The headcount doesn't vanish. It moves up the value chain and gets more expensive, which is the opposite of what most automation budgets assume.
There's a funding wrinkle worth naming. Public Law 119-21's Rural Health Transformation Program is putting real dollars into technology and AI training. The dollars are real. The people to make them work are not, unless you build them. CMS keeps expanding what hospitals are expected to do with AI and keeps declining to fund the workforce that does it. You're expected to modernize with the team you have. That only works if you reskill the team you have.
What should a hospital CFO do about this before 2030?
Three things, none of which need a capital request.
First, before you approve any AI tool, ask Stanford's question: how much downstream human work will this create, and do we have the people to do it? If the answer is no, you're not buying efficiency. You're buying a backlog.
Second, treat training as part of the technology spend, not a line you cut when margins tighten. When automation works, people get the credit. When it fails, they take the blame. The difference is almost always whether you trained them to run it.
Third, stop budgeting for fewer people and start budgeting for different people. Name the roles that shrink, name the roles that appear, and fund the bridge between them now, while you can do it on purpose instead of in a panic.
The machine isn't coming to empty your building. It's coming to change what the people in it do all day. The CFOs who see that get a sharper, higher-value team. The ones who don't will automate into a capacity crisis and spend 2029 wondering why the savings never showed up.
If you want the week's news translated this way every Sunday, I write RCM 2030 Weekly: https://www.linkedin.com/build-relation/newsletter-follow?entityUrn=7441481079762518016. The full workforce playbook is the RCM Workforce Modernization Guide on Amazon. And if your team is staring down this exact decision, that's what I help with on advisory calls.
