Two AI armies
For those of you who were hoping AI would take cost out of healthcare administration, the present situation looks dismal.
Providers are using ambient documentation and AI coding tools to get really good at capturing every billable detail of a visit. A Blue Health Intelligence analysis of Blue Cross Blue Shield claims found coding intensity climbing at hospitals as AI documentation spread. PwC's latest cost trend survey ranks AI-enabled coding among the top new pressures on 2027 costs, which plans expect to rise as much as 9% (Fortune).
Hospitals push back on that claim. The AHA calls the coding claims unsubstantiated and points to payers' own automated downcoding (AHA). In private, according to STAT, both sides concede that ambient scribes are raising coding intensity.
Payers are arming too. UnitedHealth Group expects to invest $1.5 billion in AI in 2026 (STAT), and claims review is where much of the industry's AI goes. Plans call the provider side upcoding. Providers call the payer side a denial machine.
Everyone is behaving rationally. Technology is supporting the current design. Each side is protecting its margin with the best tools it can buy. And the result is a more expensive system, and a more expensive bill for the one party without a tool of its own. So much for “disruptive technologies”… the healthcare system continues to work very well as designed.
The arms race settles in your mailbox
The fight between the hospital and the health plan ends in the patient’s mailbox and on the patient’s phone. It ends up with the patient stuck in a phone tree for an hour trying to talk to someone who knows what’s going on.
A higher-acuity code raises the allowed amount, and a patient still in their deductible or coinsurance pays a share of it. A denied or partially paid claim can stay unresolved for months while statements keep arriving. Whatever the two sides can't settle tends to roll downhill to the person with the least information. And it should not surprise us that the errors are never in the patient’s favor.
Americans paid $557 billion out of pocket for care in 2024, about 11% of all health spending, and that is before a dollar of premium (CRS). High deductibles, HSA-qualified plans and ICHRA keep moving more of the first dollar of every episode onto individuals.
The patient is no longer a bystander to the revenue cycle. The patient is a payer. And when the bills are wrong, which they often are, it’s the patient who takes the hit.
Every payer has integrity tools except one
Every other payer in healthcare checks a bill before paying it. Health plans run claims against contracts, coverage rules and clinical edits. Self-funded employers hire payment integrity vendors to audit the plan, and providers run claim scrubbers and denial management on every account.
The patient gets a PDF. Often several, including a hospital statement, a physician group bill, and an EOB that often matches neither. Reconciling them means understanding allowed amounts, network status, accumulators and coordination of benefits, then calling two phone trees to confirm.
So most people don't. When marketplace plans deny in-network claims, fewer than 1% of people appeal, even though most who do appeal succeed (PBS NewsHour). Bills follow the same pattern. People either pay whatever arrives, errors included, or they wait because they can't tell what's right, and the balance ages toward collections. One path overpays. The other creates bad debt.
Revenue cycle solutions for the patient?
The obvious response is to arm the patient. A growing set of tools now help consumers use AI to draft appeals and contest bills. They help real people, and some of them win.
But a patient bot aimed at a provider bot aimed at a payer bot is a bigger arms race, not a better system. Every automated challenge creates work for a billing office that will automate its reply. Administrative cost goes up again, with a new line item.
The goal isn't more fighting. It's accurate payment. The patient doesn't need a weapon. They need what every other payer already has: a way to know what they owe before they pay it.
What consumer payment integrity looks like
Consumer payment integrity means checking the bill on the patient's behalf, then paying it correctly. In practice it runs in four steps.
Verify. Reconcile the bill against the EOB, the claim and the plan's rules. Answer in plain language: what happened, what insurance covered, what you owe, and whether it's right.
Decide. Pay it. Wait for a claim that's still processing. Ask the provider to review a specific line, with the request drafted for you. Reimburse yourself from an HSA or HRA. Or group it with the other bills from the same episode of care.
Pay. Route the payment from the right account, with the details of what was paid and why attached to the transaction.
Remember. Keep that record with the consumer, so the next bill is easier to read and the household has one trusted history of what its care cost. Each transaction informs the next.
Much of the time the answer will be simple: this bill is right, pay it from your HSA. That's the point. Integrity isn't about finding fault; it's about knowing, then paying with confidence. Asking for a review is a step toward accurate payment, not the end state.
Why providers should want this
This only works if it's good for providers, and it is. Patient balances are the hardest dollars in the revenue cycle to collect. Statements go unread, the phones fill with "what do I owe?" calls, and accounts age into bad debt.
A patient who understands the bill and trusts the number pays it and pays it faster. The higher the trust and the more automated the better. A balance verified before payment is less likely to come back as a complaint, a call or a chargeback. An error caught early and specifically is cheaper to fix than one found in collections.
Providers have invested heavily in their side of the revenue cycle. Consumer payment integrity is the missing counterpart on the patient's side, the piece that turns a statement into a payment.
Why now
This wasn't buildable a few years ago. Three things changed.
Cost moved to the consumer. High-deductible plans, HSAs and the shift to defined contribution through ICHRA put more of the first dollar of care on individuals every year.
The data became reachable and readable. Patient access APIs and price transparency rules opened up the inputs. Modern AI can now reconcile an EOB, a bill and a plan's rules in seconds, which used to be a research project.
Money can be programmable and carry information. Payment rails can now move programmatically and can carry structured data alongside dollars. In 2025 the card networks standardized how an agent pays on a consumer's behalf with verified identity and consent.
Finish the cycle
AI in the revenue cycle isn't going away, and shouldn't go away. Accurate coding and accurate adjudication are worth automating. But a revenue cycle optimized at both ends and blind in the middle will keep getting more expensive and is not good for anyone in the long run.
The fix isn't to pick a side. It's to empower the consumer, finish the cycle, and give the patient payer the information and tools every other payer already has: the ability to automatically verify the bill, decide what to do, pay correctly, and keep the record.
If you run a benefits platform, an ICHRA administrator, a self-funded plan or a health system revenue cycle, the patient is already your counterparty. The only question is whether they show up informed. We've spent a decade at Hashed Health working on consumer-controlled health and payment records, and we're building toward this now. Reach out for more information.
