Why physicians are treating the symptoms of revenue cycle failure instead of diagnosing the disease — and how AI could change the economics of independent medical practices.
A physician would never treat a patient by examining only
the last symptom. So why do we manage medical billing by examining only the
last transaction?
The healthcare industry has become remarkably good at
processing claims. Unfortunately, it has not become equally good at preventing
the mistakes that create them.
We have electronic health records, billing platforms,
clearinghouses, coding software, dashboards, outsourcing companies, and now
artificial intelligence.
Yet physicians are still asking the same questions.
Why was this claim denied?
Why is this authorization missing?
Why has this account been sitting in accounts receivable for
ninety days?
Why is my billing company asking me to correct documentation
I completed three weeks ago?
And perhaps the most uncomfortable question:
If my practice generates all this information, why do I
still have so little visibility into my own revenue?
I believe we have been approaching the problem backward.
We keep investing in better ways to process financial
mistakes instead of building systems that prevent those mistakes from
happening.
That is not innovation. That is putting a faster engine in
an ambulance and calling it preventive medicine.
The story: The clinic that discovered its billing
department wasn't the problem
Imagine a mid-sized independent orthopedic practice.
Several physicians. A busy front desk. An electronic health
record. A billing department. An outside revenue cycle management company.
The physicians are busy seeing patients. The administrative
staff are working through their daily queues. Claims are going out.
Everything appears reasonably organized.
But the financial picture tells a different story.
Accounts receivable are increasing. Denials are recurring.
Authorization problems are delaying procedures. Documentation inconsistencies
are creating additional work.
The practice decides to investigate.
It introduces artificial intelligence to review claims,
identify coding inconsistencies, examine denial patterns, and uncover
documentation gaps.
The findings are revealing.
Some problems originate during registration. Others begin
with eligibility verification, prior authorization, clinical documentation, and
charge capture.
The billing department has been correcting mistakes that
originated long before the claim reached its desk.
The physician owner expected to discover a billing problem.
Instead, the investigation revealed an organizational
intelligence problem.
This is a hypothetical illustration, not a documented
financial recovery. But it captures a reality that many independent practices
recognize.
The last person touching a claim is not necessarily the
person responsible for the mistake.
And that distinction changes everything.
The uncomfortable truth: Billing is the autopsy of the
revenue cycle
Let me offer a different way of looking at medical billing.
The claim is not the beginning of the revenue cycle. It is
the consequence of everything that happened before it.
Consider the journey.
A patient registers. Insurance information is collected.
Eligibility is verified. An appointment is scheduled. Authorization may be
required. The physician performs an encounter. Documentation is completed.
Services are coded. Charges are captured. A claim is validated and submitted.
The payer adjudicates it. Payments are posted. Exceptions are reconciled.
Every step depends on information created in the previous
step.
One incorrect insurance entry can become a rejected claim.
One missed authorization can become a denial.
One incomplete clinical record can become a coding dispute.
One inaccurate payment posting can distort financial
reporting.
And one recurring mistake can become an expensive
organizational habit.
Yet many practices measure success primarily by what happens
at the end.
That is like judging a restaurant by how quickly it delivers
the food while ignoring whether the kitchen has the correct ingredients.
Revenue cycle management should not begin with the claim.
It should begin with the information that makes the claim possible.
The AI paradox: More intelligence does not automatically
mean better decisions
Artificial intelligence has entered healthcare with
extraordinary promises.
It can summarize clinical documentation, identify coding
inconsistencies, retrieve payer policies, categorize denials, draft appeals,
and automate repetitive administrative work.
That is genuinely useful.
But there is a problem.
The healthcare industry sometimes confuses automation with
intelligence.
Automating a broken process does not fix it. It simply
allows the broken process to operate faster.
Imagine an AI system processing thousands of claims while
relying on incomplete registration data.
Congratulations. You have automated the production of
administrative problems.
The real question is not how many tasks AI can perform.
It is whether AI can help the practice identify why those
tasks are failing.
I see three important challenges.
1. The data privacy misconception
There is a growing concern that using shared AI models
automatically gives competitors access to a clinic's proprietary billing
discoveries.
That is an oversimplification.
Enterprise AI services may provide contractual protections
against using customer data to train shared models. The actual protections
depend on the provider, service, configuration, and agreement.
An ordinary inference request does not automatically update
the model's underlying weights.
But that does not mean the practice has nothing to worry
about.
Data retention, access permissions, third-party
integrations, confidential business information, and vendor practices all
matter.
A clinic might develop a valuable payer-specific workflow
for identifying documentation gaps before submission.
That operational knowledge deserves protection.
Patient privacy protects patients. Data governance also
protects the business.
The solution is not automatically buying expensive private
AI infrastructure.
It is establishing a sensible governance framework.
Know what information enters the system. Know who can access
it. Understand how long it is retained. Review the contractual terms. Restrict
access. Maintain audit trails.
And remember that local AI is not automatically secure or
inexpensive. Hardware, maintenance, security, and staffing still cost money.
The objective is control, not technological vanity.
2. The automation cost trap
An AI agent does not necessarily stop working when the
office closes.
It can retrieve payer policies, inspect records, review
claims, draft correspondence, and communicate with other applications.
Every action may consume computing resources.
A single operation may cost very little. Thousands of
unnecessary operations can become a substantial expense.
This is the automation cost trap.
Some practices may replace a predictable human workload with
an unpredictable technology bill.
The solution is to measure outcomes rather than activity.
Track cost per claim reviewed, cost per denial prevented,
cost per appeal completed, human review time, and financial improvement
attributable to the workflow.
Local or fixed-license models may improve cost
predictability in appropriate circumstances, but they do not eliminate
operating expenses.
An AI system that saves ten minutes but costs more than
the work it replaces is not innovation. It is an expensive hobby.
3. The vendor dependency problem
Independent practices depend on an ecosystem of outside
providers.
EHR vendors. Clearinghouses. Billing companies. Cloud
platforms. AI transcription services. Patient communication applications.
Every connection introduces a dependency.
A ransomware attack or vendor outage can interrupt
operations even when the clinic's internal systems are functioning properly.
A business associate agreement is important, but signing a
contract is not the same as having a recovery plan.
Practices should establish vendor risk assessments, incident
notification requirements, backup arrangements, access restrictions, and tested
downtime procedures.
A vendor isolation protocol can help contain suspicious
activity, but indiscriminately disconnecting essential systems can interrupt
patient care.
The better approach is controlled containment.
Cybersecurity is not just about keeping intruders out. It
is about keeping the practice operational when something goes wrong.
The great divide: Physicians treat patients. Billers
treat paperwork.
Here is another uncomfortable problem.
The clinical department and the billing department often
operate like two different businesses sharing one address.
Physicians think about diagnoses, treatment, medical
necessity, and patient outcomes.
Billing professionals think about documentation
requirements, coding, modifiers, payer rules, and reimbursement.
Both groups are necessary.
Yet they frequently communicate only when something goes
wrong.
A physician completes a note.
Three days later, a biller discovers missing information.
A clarification request returns to the physician.
The physician has already moved on.
The claim waits.
The biller becomes frustrated.
The physician becomes annoyed.
And somewhere in the middle, another administrative task is
born.
We should stop treating this as a personality problem.
It is a workflow design problem.
AI-assisted natural language processing can help identify
documentation gaps and translate unstructured clinical information into
structured administrative information.
For example, a system might identify a missing documentation
element associated with a proposed service and route the issue to the
appropriate professional before claim submission.
But there is an important boundary.
AI should not invent diagnoses, manufacture medical
necessity, or insert unsupported billing codes.
The physician must retain clinical judgment. Qualified
coding professionals must review coding decisions.
Technology should improve the handoff, not manufacture the
underlying facts.
I would also introduce short, focused clinical-financial
reviews.
Not another hour-long meeting. Physicians have enough
meetings to qualify for frequent-flyer miles.
Five minutes can be enough to examine one recurring
documentation problem, identify its source, and determine whether an EHR
template or workflow should change.
The objective is to stop correcting the same mistake every
Monday.
The best billing error is the one that never reaches the
billing department.
The retrospective reporting trap
Most practices can tell you what happened last month.
They know their collections, outstanding accounts, denial
rates, and reimbursement trends.
But knowing what happened is not the same as knowing what is
about to happen.
A retrospective report might reveal that a procedure was
denied because authorization was missing.
A preventive system should identify the missing
authorization before the procedure takes place.
That is the difference between reactive billing and upstream
revenue intelligence.
Consider several practical opportunities:
- Identify
inactive insurance before an appointment.
- Flag
missing authorization before a procedure.
- Identify
incomplete documentation before coding.
- Detect
potential duplicate claims before submission.
- Prioritize
claims approaching filing deadlines.
- Identify
recurring payer-specific rejection patterns.
Predictive analytics can help identify risk patterns, but
predictions are not guarantees.
The system should explain what it identified and allow
qualified staff to investigate.
The objective is simple.
Move the intelligence closer to the decision.
Value-based care: The financial department cannot operate
in a clinical vacuum
This becomes even more important as practices participate in
value-based care arrangements.
Clinical documentation, quality measures, patient follow-up,
and financial reporting are interconnected.
A primary care physician may document a patient's chronic
conditions, review preventive care, and recommend follow-up.
The quality team needs accurate information to evaluate
performance.
The billing department needs appropriate documentation to
process claims.
The practice owner needs reliable information to understand
financial performance.
If these activities are disconnected, the organization
creates unnecessary administrative friction.
But let us be clear.
A quality measure is not simply another billing code.
Clinical care must drive documentation. Documentation must
accurately reflect care delivered. Reporting must follow the applicable program
requirements.
AI can help identify missing information and support
reporting.
It cannot manufacture quality performance.
The goal is not to turn physicians into billing clerks.
It is to make the technology understand the physician's workflow.
What the numbers tell us
The industry is moving from experimentation toward broader
AI adoption.
A November 11, 2026, Healthcare Financial Management
Association article reported that, among 95 surveyed healthcare finance
professionals:
- 27%
said their organizations were deploying AI at scale across multiple
functions.
- 53%
reported conducting AI pilots in selected areas.
- The
article also cited an Experian survey in which 69% of providers
reported that AI applications had helped reduce claim denials or improve
denial overturns.
These are survey findings, not guarantees of performance for
individual clinics.
The broader lesson is that healthcare finance leaders are
increasingly evaluating AI as an operational capability rather than an
experimental curiosity.
Prior authorization is also changing.
CMS established decision timeframes for impacted payers of
72 hours for expedited requests and seven calendar days for standard requests
under its applicable final rule. Electronic prior authorization API
requirements generally begin January 1, 2027, for impacted payers.
These changes make interoperability, workflow readiness, and
reliable information exchange increasingly important.
For physician owners, the takeaway is not to purchase every
new technology.
It is to understand where the operating environment is
changing and prepare accordingly.
Six myths about AI medical billing
Myth one: AI will eliminate billing companies.
Reality: AI can reduce repetitive work, but complex coding,
payer disputes, compliance, and exception management still require expertise.
Myth two: Shared AI automatically teaches competitors
your secrets.
Reality: Model training, inference, data retention, and
access are different issues. Vendor terms and technical configuration matter.
Myth three: Private AI is always cheaper.
Reality: Hosting, maintenance, security, and infrastructure
carry real costs.
Myth four: More automation means fewer denials.
Reality: Automation only improves outcomes when it addresses
the causes of denials.
Myth five: The billing department owns the revenue cycle.
Reality: Everyone contributing information to the patient
encounter contributes to revenue cycle performance.
Myth six: A busy billing department means a healthy
practice.
Reality: Activity is not profitability. Claims processed are
not the same as revenue collected.
A practical nine-step playbook for independent practices
Physicians do not need another technology revolution that
requires replacing everything they already own.
Start with the workflow.
Step one: Map the entire revenue cycle.
Document registration, eligibility, authorization,
documentation, coding, claims, payment posting, and reconciliation.
Identify who performs each task and where information is
transferred.
Step two: Establish a baseline.
Measure clean claim rate, denial rate, first-pass
acceptance, days in accounts receivable, net collection rate, and cost to
collect.
Use consistent definitions.
Step three: Find the recurring problem.
Identify the operational issue causing the greatest
financial or administrative burden.
Do not automate everything simultaneously.
Step four: Improve source data.
Review demographic information, insurance records, provider
identifiers, authorization information, and documentation quality.
Poor source data creates downstream problems.
Step five: Introduce targeted AI.
Start with appropriate applications such as eligibility
support, documentation checks, claim validation, or denial categorization.
Step six: Maintain human oversight.
Define which recommendations require staff approval, coding
expertise, physician clarification, or compliance review.
Step seven: Test vendor resilience.
Practice downtime procedures. Verify backups. Establish
controlled isolation protocols and recovery procedures.
Step eight: Measure results.
Compare performance against the original baseline.
Measure actual improvements rather than relying on vendor
promises.
Step nine: Scale what works.
Expand successful workflows gradually. Reevaluate anything
that increases complexity without delivering measurable value.
Tools and metrics every physician owner should understand
The technology landscape is broad, but a few categories
deserve attention.
Revenue cycle tools: EHR and practice management
systems, clearinghouses, eligibility verification, coding review, denial
management, and payment reconciliation.
Interoperability: HL7 FHIR, X12 837 claims, X12 835
remittance advice, ICD-10-CM, CPT, and HCPCS.
AI governance: NIST AI Risk Management Framework, HHS
HIPAA guidance, Coalition for Health AI resources, vendor security assessments,
and business associate agreements.
Operational metrics:
- Clean
claim rate.
- First-pass
acceptance rate.
- Initial
denial rate.
- Days
in accounts receivable.
- Net
collection rate.
- Cost
to collect.
- Authorization
turnaround.
- Documentation
rework rate.
A dashboard is useful only when someone understands what the
numbers mean and what action should follow.
Legal and ethical considerations: AI does not accept
responsibility
This deserves particular attention.
A technology vendor may promise efficiency, but the practice
must understand its own responsibilities.
Under applicable HIPAA requirements, covered entities and
business associates must address privacy, security, and contractual obligations
involving protected health information.
Business associate agreements should address permitted uses,
safeguards, incident reporting, subcontractors, and other required terms.
Vendor contracts should also address data ownership,
retention, deletion, portability, secondary use, security incidents, and
termination.
From a billing perspective, AI must never be used to create
unsupported diagnoses, unjustified modifiers, inflated service levels, or
inaccurate documentation.
From an ethical perspective, patient privacy, fairness,
transparency, and human accountability must remain central.
The objective is accurate reimbursement for appropriate
care, not reimbursement at any cost.
Three expert perspectives worth considering
The following industry perspectives offer useful lessons for
independent practices.
Nikki Harper — Healthcare Financial Management
Association
HFMA's recent revenue cycle AI discussion reinforces the
importance of identifying measurable operational problems before introducing
technology.
My takeaway: Start with the financial problem, not the
software demonstration.
Dr. Nirav Kamdar — AKASA
Industry discussions about autonomous revenue cycle
technology highlight the opportunity to automate routine work while allowing
experienced professionals to focus on more complicated situations.
My takeaway: AI should extend human capability, not
eliminate judgment.
Nick Judd and Stacey D. Bui — Cleveland Clinic and Mayo
Clinic
Their November 11 industry discussion highlights the
importance of operational implementation and workforce readiness.
My takeaway: The success of automation should be measured by
organizational improvement, not the number of automated tasks.
These are summarized perspectives, not endorsements of any
particular product or company.
The OnnX perspective: Stop thinking about billing as a
transaction
This is the philosophy behind what I am building with OnnX.
I am a physician-entrepreneur, and I believe independent
practices deserve better access to the intelligence behind their own financial
operations.
The traditional approach often creates multiple layers
between the physician and the financial information needed to understand the
practice.
An EHR stores clinical information.
A practice management system handles administrative
workflows.
A clearinghouse processes transactions.
A billing company manages claims.
Another application produces reports.
And the physician is left trying to understand how
everything connects.
That is a strange arrangement.
Imagine owning a restaurant but needing five different
companies to explain why your kitchen is losing money.
Healthcare should not have to operate that way.
My vision for OnnX is to build AI-powered revenue cycle
infrastructure that helps small and medium-sized practices improve
visibility, identify problems earlier, and make better operational decisions.
The approach is built around five dimensions:
Process reliability: Are the necessary workflows
being completed consistently?
Visibility: Can the practice understand where
financial and administrative problems exist?
Prevention: Can errors be identified before they
become denials?
Resolution: Can outstanding problems be prioritized
and addressed?
Data integrity: Is the information supporting each
transaction accurate and complete?
These dimensions work together.
Improving one without understanding the others can produce
disappointing results.
For example, faster claim submission is not necessarily an
improvement if coding accuracy declines.
A sophisticated denial dashboard is not enough if the
practice cannot identify why the same errors continue.
And an AI agent is not valuable simply because it can
perform thousands of tasks.
The real opportunity is connecting information from the
beginning of the patient journey to the final payment.
What does eliminating middlemen actually mean?
I use that phrase deliberately, but not to suggest that
every external billing company is unnecessary.
Experienced billing professionals provide real value.
The issue is unnecessary dependency, redundant handoffs,
fragmented information, and limited financial visibility.
Physicians should have access to their own operational
intelligence, whether they choose to manage billing internally or work with
outside partners.
Eliminate unnecessary intermediaries. Preserve valuable
expertise. Give the practice greater control.
That is the distinction.
Future outlook: The next generation of medical billing
Over the next several years, I expect the industry to evolve
in six important directions.
First, preventive intelligence will become more
important than retrospective reporting.
Second, interoperability will increasingly determine
how effectively healthcare systems exchange information.
Third, human-supervised automation will become a more
practical model than unrestricted autonomous decision-making.
Fourth, data governance will become a competitive
requirement rather than a technical afterthought.
Fifth, revenue intelligence will become more
important to independent physician owners who want greater visibility into
their financial operations.
Finally, technology providers will increasingly be judged by
whether they improve actual practice performance rather than simply offering
more features.
The winners will not necessarily have the largest models or
the most AI agents.
They will understand the workflow, protect the information,
and solve problems that matter.
Final thoughts: Diagnose the system before prescribing
another solution
I began with a simple question.
Why would a physician investigate only the last symptom when
the underlying disease may have started much earlier?
Medical billing deserves the same analytical discipline we
apply to medicine.
Identify the cause. Examine the evidence. Understand the
system. Intervene at the right point. Measure the result.
Stop treating denials as isolated events. Start
understanding the processes that create them.
Stop buying automation for the sake of automation. Start
demanding measurable intelligence.
Stop allowing the people who generate revenue cycle
information to operate in disconnected silos.
Independent practices do not need more administrative
complexity.
They need better information, smarter workflows, and greater
control over the financial infrastructure supporting patient care.
That is the opportunity I see in AI-powered medical billing.
And that is the future I want to help build.
Get involved: Join the conversation
Let me leave you with three questions.
1. The provocative question: If your billing
department disappeared tomorrow, would you actually know where your practice
was losing revenue?
2. Share your experience: What creates the biggest
financial headache in your practice: denials, prior authorization, coding,
documentation, or delayed payments? Leave a comment. I want to hear from
physicians and practice owners.
3. Help spread the conversation: If this perspective
resonates, consider reposting it. Another physician or clinic owner may be
struggling with exactly the same problem.
The future of medical billing should not be about adding
another administrative layer.
It should be about removing unnecessary complexity and
giving physicians greater control.
Let's build a smarter revenue cycle together.
About the Author
Dr. Daniel Cham, MD, MS is a physician, healthcare
strategist, and entrepreneur focused on healthcare technology, medical practice
management, and revenue cycle innovation.
As founder of OnnX, he is working toward a more
intelligent approach to medical billing for independent and
small-to-medium-sized medical practices.
His interests include AI-powered revenue cycle
infrastructure, healthcare interoperability, data integrity, operational
efficiency, and reducing unnecessary administrative complexity.
Disclaimer
This article is intended for general educational and
informational purposes. It does not constitute legal, medical, coding,
cybersecurity, or financial advice. Readers should consult appropriately
qualified professionals regarding their specific circumstances.
Continue the Conversation
Explore practical perspectives, emerging technologies, and
thoughtful commentary on the future of medicine, healthcare operations, and
innovation.
Explore more:
Connect: LinkedIn | Website | YouTube| Spotify
Knowledge creates opportunity. Better information creates
better decisions. Start here.
PS: Visit the Featured section of my LinkedIn profile
for a free resource. No signup required.
If this perspective resonates, consider reposting to help
physicians and clinic owners rethink how billing impacts their practices.
References
Healthcare
Doesn’t Have an AI Problem. It Has an Information Problem. And Your Billing
Department Is Paying for It. — Explores how fragmented, inaccurate, and
incomplete healthcare information limits the effectiveness of artificial
intelligence.
Your
Medical Billing Problem May Not Be a Billing Problem: The Hidden Cost of
Creating Work You Should Never Have to Do — Examines how upstream
administrative failures create preventable denials, rework, and revenue-cycle
inefficiencies.
Your
AI Might Be Your Best Employee. Or Your Most Expensive Competitor. —
Examines AI data ownership, vendor dependency, privacy, and the strategic value
of proprietary operational intelligence.
AI-Driven
Medical Billing Explained | Faster Reimbursements, Fewer Denials —
Introduces AI-powered medical billing, automation, and opportunities to improve
revenue-cycle operations.
What
Is AI in Healthcare Revenue Cycle? — Healthcare Financial Management
Association — Reviews AI applications in healthcare revenue-cycle
management, including automation, denial prevention, and industry adoption.
AI
in Revenue Cycle: From Point Solutions to Enterprise Value — PwC — Explains
why healthcare organizations should evaluate AI through end-to-end workflow
improvement and measurable financial outcomes rather than isolated automation.
Hashtags
#OnnX #MedicalBilling #HealthcareAI #RevenueCycleManagement
#RevenueIntelligence #PhysicianEntrepreneur #MedicalPracticeManagement
#HealthcareTechnology #MedicalCoding #HealthcareInnovation
#IndependentPractices #DigitalHealth #PhysicianLeadership #ArtificialIntelligence