What a 4-year-old boy, two physician parents, and a nurse driving four hours through traffic can teach us about the hidden problems that begin long before a claim is submitted
“This is about creating a future healthcare system that
is increasingly tech-enabled and always people-centred.” — Professor
Alastair Denniston and Professor Henrietta Hughes, National Commission
into the Regulation of AI in Healthcare
Four-year-old Oliver has Duchenne muscular dystrophy.
His parents, Dr. Alec James and Dr. Audreanna James,
are family physicians in Hampshire County, West Virginia.
They understand medicine.
They understand diagnoses.
They understand treatment.
They understand medical terminology.
They know which questions to ask and whom to ask.
And then their own son became the patient.
Suddenly, their medical knowledge could not make the
healthcare system simple.
Oliver takes Exondys 51, a treatment approved by the
FDA a decade ago. But approval was not the end of his journey.
It was the beginning of another one.
Coverage.
Prior authorization.
Specialty pharmacy.
Clinical coordination.
Treatment delivery.
And, once a week, a nurse drives from Chantilly, Virginia,
across the state line to reach Oliver.
Depending on traffic, the trip can take two hours each
way.
Four hours in a car.
So a treatment created through years of science, research,
clinical trials, regulation, investment, and medical expertise ultimately
reaches a 4-year-old boy because one person gets in a car and shows up at
his kitchen counter.
That is what access looks like in real life.
Not on a PowerPoint.
Not in a press release.
Not in a dashboard.
At a kitchen counter.
And that is where this story becomes unexpectedly relevant
to medical billing.
Because healthcare has a peculiar habit.
We celebrate the beginning.
We celebrate the breakthrough.
We celebrate the approval.
We celebrate the technology.
And then we act surprised when the last mile doesn't work.
Maybe the last mile is actually the whole game.
The Claim Is Where We Notice the Problem
Let's move from Oliver's kitchen to a physician practice.
A patient arrives.
The physician evaluates the patient.
The physician documents the encounter.
Someone selects codes.
A claim is generated.
The claim is submitted.
Then comes the message nobody wants:
Denied.
The billing team opens the claim.
Someone reviews the chart.
Someone checks eligibility.
Someone searches for an authorization.
Someone sends a query.
Someone calls the payer.
Someone waits.
Someone calls again.
Someone resubmits.
And eventually someone asks:
“What happened?”
That question may be the most expensive question in the
revenue cycle.
Because by the time you're asking it, the original context
may already be gone.
The patient left days ago.
The physician has seen dozens of other patients.
The scheduler has moved on.
The authorization team is working on another case.
The biller is staring at a claim that appears to have
materialized from another dimension.
And everybody is trying to reconstruct the past.
Healthcare calls this revenue-cycle management.
Sometimes it feels more like archaeological excavation.
Here Is the Contrarian Part
I don't think the future of medical billing is primarily
about becoming better at fixing claims.
I think the bigger opportunity is becoming better at preventing
claims from becoming problems.
That sounds obvious.
It isn't.
Because the healthcare industry has spent years optimizing
the downstream machinery.
Claims systems.
Clearinghouses.
Denial management.
Appeals.
A/R workflows.
Payment posting.
Analytics.
Work queues.
Automation.
And all of those things matter.
But there is an uncomfortable question hiding underneath
them:
What if we have become extremely good at managing
problems we should have detected earlier?
That is a very different problem.
The Denial Is Not the Disease
A denial is an event.
It is not necessarily the root cause.
Think about a smoke alarm.
When the alarm goes off, you have learned something
important.
But you haven't necessarily learned where the fire started.
The same is true of a denial.
The denial tells you:
Something went wrong.
It doesn't automatically tell you:
Why.
And it certainly doesn't guarantee that the failure began in
billing.
It might have begun at registration.
Eligibility.
Scheduling.
Authorization.
Documentation.
Coding.
Charge capture.
Payer configuration.
Workflow design.
Or simply because information existed somewhere but wasn't
available when a decision had to be made.
The claim is often where the problem becomes visible.
Not where it becomes real.
Healthcare Doesn't Have a Data Shortage
This may be one of healthcare's strangest contradictions.
We have enormous amounts of data.
Clinical notes.
Lab results.
Imaging.
Diagnoses.
Procedure codes.
Eligibility information.
Authorizations.
Claims.
Remittance advice.
Payer rules.
Scheduling data.
Patient demographics.
Financial information.
And yet people constantly ask:
“Where is that information?”
The problem isn't necessarily missing data.
It is often disconnected context.
Information exists.
But it exists:
in another system,
in another workflow,
with another person,
under another label,
or five clicks away from where the decision is being made.
That's not a data problem.
That's a decision-timing problem.
The Information Gets Thinner as It Travels
Think about the clinical encounter.
A physician sees a human being.
Not a code.
Not a claim.
A person.
The physician sees history.
Symptoms.
Risk.
Prior treatment.
Clinical reasoning.
Uncertainty.
Context.
A decision.
Then that rich clinical encounter has to become structured
information.
Documentation.
Then coding.
Then a claim.
Then payer adjudication.
Each translation is necessary.
But every translation creates an opportunity for context
loss.
By the time the payer sees the claim, the original clinical
story may have been compressed into a relatively small set of structured
fields.
Efficient?
Yes.
Complete?
Not always.
And that distinction matters.
The Biller May Be Solving the Wrong Problem
Let's defend billing teams for a moment.
Your biller may be excellent.
Your coder may be excellent.
Your revenue-cycle company may be excellent.
Your denial specialist may be excellent.
And you can still have expensive revenue leakage.
Why?
Because downstream experts cannot magically recover
information that was never captured, never connected, or never surfaced at the
right moment.
Giving an excellent biller a bad input does not transform
the input into a good one.
It simply creates an excellent biller with a headache.
And probably a spreadsheet.
Healthcare has enough spreadsheets.
The Hidden Cost of “Again”
One of the most expensive words in healthcare administration
may be:
Again.
Check it again.
Send it again.
Document it again.
Verify it again.
Explain it again.
Submit it again.
Appeal it again.
Call them again.
Review it again.
Every “again” represents rework.
And rework is expensive.
Not just because of wages.
Because of attention.
A skilled employee spending 20 minutes reconstructing a
preventable problem isn't spending those 20 minutes doing something else.
Multiply that across hundreds or thousands of encounters.
Now the problem has a financial dimension.
And an operational one.
And a human one.
Oliver's Story Gives Us a Different Definition of Access
The James family makes an important observation in their
account.
They have advantages that many families don't.
They understand medical language.
They know what questions to ask.
They know whom to contact.
And despite all of that, navigating the system has still
been difficult.
That should make us pause.
If two physicians can find the healthcare system difficult
to navigate for their own child, what does the experience look like for someone
without that knowledge?
That is not an argument against healthcare.
It is an argument for making healthcare easier to
navigate.
And that principle applies to billing too.
If a physician practice requires an expert detective to
reconstruct what happened after a claim fails, the workflow may be telling us
something.
Maybe the information arrived too late.
Maybe it arrived in the wrong place.
Maybe it was never connected.
Maybe the system assumed the human would figure it out.
Humans are remarkably good at figuring things out.
That doesn't mean we should make them do it repeatedly.
The Last Mile Is Not Administrative
We often describe the final steps of healthcare as
administrative.
That word is convenient.
It is also misleading.
The last mile can determine whether the patient actually
receives something.
In Oliver's case, the nurse's journey is not merely
logistics.
It is part of access.
The James family's point is powerful because the treatment
exists only in theory if the chain between approval and administration breaks.
The same concept exists in revenue cycle.
A clinically appropriate service can be delivered.
The documentation can exist.
The coding can be appropriate.
And yet the practice may still struggle to get paid
accurately if the information chain breaks.
Different problem.
Same systems principle.
The chain matters.
The AI Industry Is Asking the Wrong Question Too
Healthcare AI is having its own version of this problem.
The question often becomes:
“What can AI automate?”
I think a better question is:
“What decision can AI help a human make earlier and
better?”
That is much less glamorous.
Which is precisely why I like it.
The UK's National Commission into the Regulation of AI in
Healthcare published recommendations this month emphasizing that AI should be
people-centred, safe, trusted, and subject to human oversight. The Commission's
report says AI should augment rather than replace healthcare professionals and
that technologies need to demonstrate safety, effectiveness, and benefit.
That philosophy translates surprisingly well to revenue
cycle.
AI does not need to become the boss.
It can become the early-warning system.
Imagine a Different Billing Workflow
Instead of:
Encounter → Claim → Denial → Investigation
imagine:
Encounter → Understand → Detect → Decide → Act → Claim
The difference is not necessarily more automation.
It is earlier intelligence.
The system understands the available context.
It detects something unusual.
It surfaces the issue.
A human decides what to do.
The workflow proceeds.
The claim is generated with fewer surprises.
That is a fundamentally different philosophy.
Don't Let AI Become a Very Fast Wrong Answer
Here's another unpopular opinion:
Healthcare does not need AI that confidently makes more
mistakes.
It needs AI that understands when to slow down.
Suppose an AI system detects a possible documentation-coding
mismatch.
A reckless system says:
“Change the code.”
A better system says:
“Potential mismatch detected. Here is the documentation
supporting the current choice and the information that may warrant review.”
The second system preserves human judgment.
It also creates an audit trail.
And it avoids pretending that uncertainty doesn't exist.
That matters.
Especially in healthcare.
Automate the Predictable. Escalate the Uncertain.
This should be one of the basic principles of healthcare AI.
Automate what is repetitive.
Surface what is unusual.
Escalate what is uncertain.
Preserve what is clinically important.
Keep humans accountable.
The goal is not:
Human versus AI.
The goal is:
Human judgment with better timing and better information.
A Practice Experiment That Costs Almost Nothing
Here is something a small practice can do next week.
Take 25 recent problem claims.
Don't start by fixing them.
Start by tracing them backward.
For every claim, ask:
Where did the problem first appear?
Not where it was discovered.
Where did it begin?
Create a simple list:
- Registration
- Eligibility
- Authorization
- Documentation
- Coding
- Charge
capture
- Payer
- Workflow
- Other
Then ask a second question:
When could we first have detected it?
That question changes everything.
The “Earlier” Test
For each problem, ask:
Could we have known 30 days earlier?
Maybe not.
Could we have known one week earlier?
Maybe.
Could we have known before the encounter?
Sometimes.
Could we have known before submission?
Often.
Now you have something useful.
You have identified a decision window.
That is much more actionable than a denial report.
Measure the Right Thing
Denial rate matters.
But it shouldn't become the entire story.
Track:
Manual touches per claim
Rework minutes per encounter
Documentation queries
Coding queries
Authorization-related failures
Claims corrected before submission
Clean-claim rate
Days from encounter to submission
A/R aging
Underpayment variance
Appeal recovery
And perhaps the most interesting:
Problems detected before submission
That's a leading indicator.
It asks whether your system is getting smarter before
the payer tells you something went wrong.
A Denial Dashboard Can Lie Without Lying
This is an important distinction.
A dashboard can report accurate numbers and still give
management the wrong impression.
Suppose denials fall.
Everyone celebrates.
But staff workload rises because underpayments increase.
Or payment timing deteriorates.
Or the remaining denials become harder to resolve.
Or the practice is spending more time on fewer but more
expensive problems.
The metric wasn't wrong.
The interpretation was incomplete.
Healthcare operations need causal curiosity.
When a number changes, ask:
Why?
The Five Whys of Revenue Cycle
A denial appears.
Why?
Missing authorization.
Why?
Authorization was not identified.
Why?
The scheduling workflow did not surface the requirement.
Why?
The payer rule was not connected to the service
configuration.
Why?
The system treated authorization as a downstream billing
task.
Now we have learned something.
The problem wasn't really:
“Biller missed authorization.”
It was:
“The workflow discovered an authorization requirement too
late.”
That is a very different intervention.
Stop Calling Everything a Billing Problem
Here's another provocative thought.
Some billing problems aren't billing problems.
They're:
registration problems,
scheduling problems,
clinical documentation problems,
authorization problems,
workflow problems,
information architecture problems,
or decision-timing problems.
Billing simply becomes the department where the consequences
finally become visible.
That distinction matters because organizations tend to send
problems to the department where they are discovered.
Not necessarily the department where they originated.
The “Throw It Over the Wall” Model
Healthcare has an unfortunate organizational habit.
Registration finishes its part.
Then it throws the information over the wall.
Scheduling finishes its part.
Over the wall.
Clinical documentation.
Over the wall.
Coding.
Over the wall.
Billing.
Over the wall.
Payer.
Over the wall.
Eventually someone discovers the wall has accumulated a pile
of problems.
Then we hire another person to clean up the pile.
Maybe the future isn't a better cleanup crew.
Maybe it's fewer walls.
What OnnX Is Exploring
This is the idea behind OnnX.
Not another traditional billing service.
Not simply another dashboard.
Not “AI that does billing.”
The underlying question is more fundamental:
What happens when billing intelligence moves upstream?
OnnX explores an AI-powered revenue-cycle model for
physician-owned practices in which patient, clinical, documentation, coding,
and workflow context can inform decisions before the claim becomes a
downstream problem.
The conceptual loop is:
Understand → Detect → Decide → Act → Learn
Understand the practice.
Understand the encounter.
Detect potential problems.
Surface relevant options.
Keep human judgment in the loop.
Learn from actual outcomes.
Then improve.
The objective isn't to remove humans.
It is to give humans better information before they have
to clean something up.
This Is Not a Magic AI Story
There is an important caveat.
AI will not eliminate payer complexity.
It will not eliminate exceptions.
It will not eliminate human judgment.
It will not make healthcare simple.
And it certainly will not make every claim payable.
That would be a fantasy.
The opportunity is narrower.
And more practical.
Reduce avoidable uncertainty.
Surface problems earlier.
Preserve context.
Reduce unnecessary rework.
That's enough.
The Compliance Line Must Stay Bright
Any technology operating in medical billing has to respect
the boundary between accurate representation and financial manipulation.
AI should not invent documentation.
It should not encourage unsupported coding.
It should not alter the clinical record to increase
reimbursement.
It should not turn payer prediction into clinical
decision-making.
And it should not hide uncertainty.
The fundamental principle remains:
Document what happened.
Code what is supported.
Bill what was actually provided.
Technology should make that process more accurate and
transparent—not more aggressive.
The Human-in-the-Loop Is Not a Weakness
Some technology companies talk about human intervention as
if it means the AI isn't good enough.
I disagree.
In healthcare, human oversight can be a feature.
If the system says:
“I found something worth reviewing.”
That's useful.
If it says:
“I know exactly what to do.”
when the evidence is ambiguous, that's dangerous.
Good systems should know the difference.
The Real Opportunity: Decision Infrastructure
Maybe the future of revenue cycle isn't another billing
application.
Maybe it's something closer to decision infrastructure.
A layer that helps connect:
Patient context.
Clinical context.
Documentation.
Coding.
Payer requirements.
Workflow.
Historical outcomes.
And human judgment.
Not replacing the existing ecosystem overnight.
Making the ecosystem smarter about when and where
decisions happen.
That's a different category.
Three Questions for Every AI Healthcare Startup
Before building another AI feature, ask:
1. What decision are we improving?
If the answer is unclear, the feature may be technology
looking for a problem.
2. When does that decision currently happen?
Timing is everything.
3. What information was unavailable when the decision was
made?
That question may reveal the real product opportunity.
Three Questions for Every Physician Owner
Ask your team:
What problem do we keep fixing?
Where does it actually begin?
How much earlier could we detect it?
Then ask the question nobody likes:
Why are we still fixing it manually?
Not because someone is doing a bad job.
Because the system may be designed to require it.
What Oliver Teaches Us About Technology
Oliver's story is not a technology story.
That is precisely why it matters to technology.
The healthcare system ultimately succeeds when something
tangible happens for a human being.
A treatment reaches a child.
A patient gets an appointment.
A physician has the information needed to make a decision.
A claim accurately represents the care delivered.
A practice gets paid for legitimate work.
Technology is the machinery.
The human outcome is the product.
Healthcare's Strange Obsession With the Middle
We tend to focus on the middle.
The software.
The workflow.
The claim.
The code.
The dashboard.
The AI.
But patients care about the beginning and the end.
Can I get the care?
Did the care work?
Can the physician continue providing it?
And physician owners care about something equally
fundamental:
Can the practice sustain itself while providing that
care?
The middle exists to connect those outcomes.
We shouldn't confuse the machinery with the mission.
The Last Mile Is Where Strategy Becomes Reality
The James family could have stopped at FDA approval.
They didn't.
Because approval isn't access.
A prescription isn't treatment.
A claim isn't payment.
Documentation isn't necessarily accurate coding.
A code isn't necessarily a clean claim.
A clean claim isn't necessarily cash.
Every step has another step behind it.
And every step creates another opportunity for information
to be lost.
That is why upstream thinking matters.
The Most Interesting Question in Medical Billing
It isn't:
“How do we get more claims paid?”
It is:
“What information would have allowed us to make a better
decision earlier?”
That question moves the conversation from transactions to
systems.
From correction to prevention.
From reactive billing to upstream intelligence.
From more work to potentially less work.
And from:
“Who caused this?”
to:
“Where did the system lose the information?”
That is a much more productive question.
A 30-Day Upstream Billing Experiment
You don't need a million-dollar transformation.
Week 1: Find the friction
Choose 25–50 recent claims that required unusual
intervention.
Week 2: Trace backward
Identify the earliest point where the problem appeared.
Week 3: Move detection
Pick one problem and move its detection earlier.
Week 4: Measure
Compare:
- Manual
touches
- Rework
time
- Corrections
before submission
- Denials
- Days
to clean submission
- Staff
satisfaction
Then ask:
Did we actually make the work easier?
If yes, expand.
If no, learn.
That is innovation without the theater.
The Future May Be Quiet
The most valuable AI in healthcare may not be the AI that
produces the most impressive demo.
It may be the AI that quietly prevents a problem.
No dramatic dashboard.
No fireworks.
No robot voice.
Just:
“This may need attention before you submit.”
And then the person fixes it.
Five minutes.
Done.
Nobody writes a press release.
That's okay.
The absence of a problem rarely makes headlines.
But practices notice.
Maybe the Best AI Is Boring
I will make one final contrarian prediction about product
design—not about election outcomes, politics, or anything else.
The best healthcare AI may eventually become remarkably
boring.
It won't ask clinicians to learn an entirely new universe.
It won't demand ten new dashboards.
It won't create another inbox.
It won't produce 47 alerts.
It will simply appear at the right moment with the right
information.
Then disappear.
That sounds almost disappointingly simple.
Good.
Healthcare has enough complicated software.
The Lesson From Oliver's Kitchen Counter
Return to that image.
A 4-year-old boy.
A nurse.
A treatment.
A kitchen counter.
The nurse drove four hours through traffic because access
does not happen automatically.
The James family understands this personally.
They wrote that a therapy only changes a child's life if the
child can actually receive it. They also described how FDA approval is only one
link in a much longer chain involving coverage, prior authorization, specialty
pharmacy, clinicians, and treatment delivery.
That is the lesson.
The chain matters.
In healthcare delivery.
In healthcare technology.
And in medical billing.
A claim is also a link in a chain.
It is not the beginning.
And it is certainly not the whole story.
Final Thought
Healthcare does not need another system that proudly
announces:
“We found 1,842 problems.”
It needs systems that help answer:
“Which problems could we have prevented?”
Then:
“How much earlier could we have known?”
Then:
“What information would have changed the decision?”
That is where the opportunity becomes interesting.
Because the future of medical billing may not belong to the
company that processes the most claims.
It may belong to whoever helps practices make better
decisions before the claim becomes a problem.
Oliver's story makes that idea human.
One boy.
Two physician parents.
One nurse.
One long drive.
One treatment.
And an enormous healthcare system working—or failing to
work—between them.
The lesson isn't that technology should replace people.
It is almost the opposite.
Technology should make it easier for people to do the
part that only people can do.
And maybe that's the real meaning of a people-centred
healthcare system.
Not less technology.
Better technology in the right place.
The Question I Want to Leave You With
What is the one “billing problem” your practice keeps
fixing that actually began somewhere upstream?
Registration?
Eligibility?
Authorization?
Documentation?
Coding?
Workflow?
Something else?
I would genuinely like to hear the answer.
Because the most interesting revenue-cycle problems aren't
always the ones appearing on the denial report.
Sometimes they're hiding three steps earlier.
If this perspective made you rethink where a billing
problem actually begins, share it with another physician, practice owner,
administrator, or biller.
The conversation is more valuable when more people join it.
About the Author
Dr. Daniel Cham is a physician, medical consultant,
and healthcare entrepreneur working at the intersection of medical
technology, healthcare operations, and medical billing.
As founder of OnnX, he explores how AI can help
physician-owned practices identify revenue-cycle problems earlier, preserve
clinical context, reduce avoidable rework, and support better decisions before
claims become downstream problems.
His central question is simple:
What if the claim is where the problem becomes
visible—not where it begins?
Connect with Dr.
Daniel Cham on LinkedIn
Disclaimer
This article is for general educational and informational
purposes only. It does not constitute medical, legal, coding, compliance,
reimbursement, financial, or other professional advice.
Healthcare requirements vary by payer, specialty,
contract, jurisdiction, and individual circumstances. Practices should obtain
appropriate professional guidance for specific clinical, legal, coding,
compliance, privacy, and reimbursement questions.
Continue the Conversation
LinkedIn: Connect with Dr.
Daniel Cham
Website: Dr.
Daniel Cham
Spotify: Listen to the
podcast
YouTube: Watch
on YouTube
X: Follow on X
Facebook: Follow on
Facebook
Free Resource
Visit the Featured section of my LinkedIn profile for
a free resource designed to help physicians and practice leaders think
differently about the relationship between clinical workflow, documentation,
coding, billing, and revenue.
No signup required.
Start with the information you already have. Then ask
what you could have known earlier.
References
1. Drs. Alec James and Audreanna James — “Ten Years After
the First Duchenne Drug, Our Son Shows Why Access Matters,” DC Journal,
September 18, 2026.
Read
the original story
2. National Commission into the Regulation of AI in
Healthcare — Recommendations for a Future Regulatory Framework, UK Government,
September 10, 2026.
Read
the Commission report
3. Medicines and Healthcare products Regulatory Agency —
“Independent Commission led by NHS doctors sets out blueprint to accelerate
safe AI adoption in healthcare,” September 10, 2026.
Read
the MHRA announcement
Hashtags
#MedicalBilling #HealthcareRevenueCycle #RCM #HealthcareAI
#MedicalPractice #PhysicianPractice #HealthcareOperations #MedicalCoding
#ClinicalDocumentation #HealthTech #HealthcareInnovation #PracticeManagement
#PhysicianEntrepreneur #DigitalHealth #HealthcareTechnology
#RevenueCycleManagement #PhysicianOwnedPractice #AIinHealthcare #OnnX
No comments:
Post a Comment