The uncomfortable truth about medical billing, healthcare data, artificial intelligence, and why independent medical practices keep paying for problems that should have been prevented upstream.

“Better is possible. It does not take genius. It takes diligence. It takes moral clarity. It takes ingenuity. And above all, it takes a willingness to try.” — Dr. Atul Gawande, surgeon and author

The Broken Wrist That Broke the Billing System

A patient walks into a medical practice with a fractured wrist.

The physician examines the patient, confirms the diagnosis, applies a cast, documents the encounter, and completes the treatment.

The biller enters the appropriate procedure code.

The claim goes out.

The clinic gets paid.

That is how it should work.

Except it doesn’t always work that way.

The insurance information might be outdated. The eligibility check might never have happened. The diagnosis might be buried in a paragraph instead of entered into the appropriate structured field. An authorization might be missing. A modifier might have been overlooked.

Now the biller is investigating.

The physician is being interrupted.

The office manager is calling the insurance company.

Someone is searching through old documentation.

And the claim that should have taken minutes to prepare becomes a small administrative crime scene.

Here is the interesting part.

The fracture was treated correctly. The billing problem may have started before the patient ever entered the examination room.

Think about that.

We can diagnose a broken bone, take an X-ray, apply a cast, and send the patient home. Yet the administrative machinery responsible for getting the practice paid can struggle to establish whether the insurance information was correct.

We have sophisticated medical technology.

We have electronic health records.

We have cloud computing.

We have artificial intelligence.

But sometimes the person responsible for collecting payment still needs to ask Susan what happened to the authorization.

If Susan is on vacation, apparently the technology is on vacation, too.

That is not a technology strategy.

That is institutional amnesia with a software subscription.

The problem with healthcare automation is that we keep making the wrong things faster.

And independent physicians are paying the price.

The Real Healthcare AI Debate Nobody Wants to Have

The healthcare industry is excited about artificial intelligence.

Every week, someone announces another AI platform.

AI coding. AI documentation. AI eligibility. AI prior authorization. AI denial prediction. AI revenue cycle management.

The promises are impressive.

Fewer employees. Faster claims. Lower administrative costs. Higher collections.

But there is a question that deserves more attention:

What happens when you automate a process that is already working with inaccurate information?

You do not necessarily eliminate the problem.

You may simply distribute it faster.

Imagine giving a medical biller an AI system that can process 10,000 claims instead of 1,000.

Sounds wonderful.

But what if the patient identities are wrong?

What if the insurance information is outdated?

What if the clinical documentation is incomplete?

What if the system interprets a physician’s narrative incorrectly?

Congratulations. You have successfully automated confusion.

The healthcare industry often approaches AI as if the primary problem is insufficient computing power.

I would argue that the more fundamental problem is insufficient information discipline.

Healthcare does not need more intelligence applied to information it cannot trust. It needs better information before intelligence is applied.

This is where the debate becomes interesting.

Should medical practices first build standardized, structured data systems before deploying AI?

Or should they use AI immediately to clean the incomplete and inconsistent information they already have?

Both arguments have merit.

But neither tells the whole story.

For independent practices, waiting for perfect infrastructure is unrealistic.

Deploying AI without validation is irresponsible.

The answer is not choosing one technology over another.

It is understanding what each technology should be allowed to do.

The First Mistake: Confusing More Data With Better Data

Healthcare produces enormous amounts of information.

Physicians document symptoms, examination findings, diagnoses, treatment plans, prescriptions, and follow-up instructions.

Laboratories generate results.

Imaging departments produce reports.

Insurance companies generate eligibility responses, authorization requirements, remittance information, and claim decisions.

Electronic health records collect much of this information.

Yet collecting information and making it useful are two different things.

Consider a typical medical encounter.

The physician documents the patient’s condition in a SOAP note.

The biller needs the appropriate diagnosis and procedure information.

The insurance company needs the information required to adjudicate the claim.

The practice management system needs the information to support the revenue cycle.

The problem is that these systems may not interpret or exchange information consistently.

One system contains a narrative.

Another expects a structured field.

A third expects a specific code.

The information exists, but the connections are missing.

This is a crucial distinction.

Data availability is not data usability. Data usability is not data accuracy. And data accuracy is not automatically clinical validity.

These are different problems.

Healthcare organizations frequently treat them as if they were interchangeable.

They are not.

The Two Competing Schools of Thought

Argument One: Standardize the Data Before Introducing AI

This position begins with a reasonable assumption.

A computer system needs consistent information to perform reliable operations.

Healthcare data comes from different systems, formats, and terminology standards.

The same patient may appear under different names.

The same clinical concept may be represented differently.

The same insurance requirement may be stored in a different location depending on the software.

One solution is FHIR, or Fast Healthcare Interoperability Resources, developed by HL7.

FHIR provides a standardized framework for exchanging healthcare information.

It defines resources representing information such as:

• Patients.
• Conditions.
• Observations.
• Medications.
• Procedures.
• Encounters.
• Diagnostic reports.

FHIR can improve interoperability by giving different systems a common way to represent and exchange information.

But there is a catch.

Standardizing incorrect information does not make it correct.

If a patient’s date of birth is wrong, a perfectly formatted FHIR resource can still contain the wrong date of birth.

If the physician never documented a required clinical finding, FHIR cannot magically create it.

If an insurance policy is inactive, a standardized data structure cannot make it active.

Standards improve consistency.

They do not replace verification.

Another important concept is policy as code.

Instead of asking employees to remember every payer requirement, organizations can translate defined rules into machine-readable instructions.

For example, a system can check whether:

• Required patient information is present.
• Insurance coverage has been verified for the date of service.
• A required authorization is recorded.
• A claim contains required information.
• A known payer-specific requirement has been addressed.
• A particular coding combination requires review.

The benefit is straightforward.

Instead of discovering every problem after submission, the system can identify many problems before submission.

That is meaningful progress.

But there is a potential downside.

What happens when the system encounters a field that is blank because the physician never entered the information?

The system can stop the workflow.

But stopping a workflow is not the same as solving the problem.

The biller still needs the information.

The physician still needs to be contacted.

The patient may still be waiting.

A system that identifies every problem but cannot help resolve any of them is not intelligent automation. It is an electronic hall monitor.

Argument Two: Use AI to Clean the Data That Already Exists

The second position is more practical for many independent practices.

Most small and medium-sized clinics cannot afford to replace their entire technology infrastructure.

They already have years of clinical documentation, scanned records, inconsistent patient information, and fragmented administrative workflows.

Waiting for a perfect data environment means waiting indefinitely.

This is where targeted AI becomes useful.

Natural language processing can identify information in clinical notes.

Document intelligence can extract information from insurance cards and scanned forms.

Entity-resolution algorithms can identify potentially duplicated patient records.

Machine learning can flag unusual patterns and inconsistent information.

AI can also help organize tasks, identify missing documentation, and direct exceptions to the appropriate employee.

Consider a physician’s SOAP note.

The physician documents the patient’s symptoms, examination, assessment, and treatment plan.

The information exists.

But it may not be available in the structured format another system needs.

Natural language processing can extract candidate information and organize it into structured fields.

This can reduce repetitive manual work.

However, there is an important distinction:

Extracting information is not the same as verifying information.

An algorithm can identify a diagnosis mentioned in a note.

It cannot automatically determine that the diagnosis is appropriate for every billing purpose.

It can identify a missing blood pressure.

It cannot establish what the patient’s blood pressure actually was.

It can suggest that two records belong to the same person.

It cannot safely merge every similar-looking record without considering the consequences.

AI should help people find information.

It should not manufacture reality.

My Position: Stop Treating This as an Either-Or Decision

The two approaches are not enemies.

They solve different parts of the problem.

AI helps organize messy information.

Standardization makes information more consistent.

Validation checks whether defined requirements are satisfied.

Human review resolves uncertainty.

Audit trails preserve accountability.

The sequence should be practical:

Extract. Organize. Validate. Reconcile. Escalate. Automate.

Start with the information already available.

Use AI to identify what is missing or inconsistent.

Apply deterministic rules to what can be verified.

Ask humans to resolve ambiguous or consequential decisions.

Preserve the source and history of important information.

Then gradually improve the underlying data infrastructure.

Independent practices do not need to build a technology empire before benefiting from automation.

They need a sensible starting point.

Five Healthcare Data Problems Hiding Inside Medical Billing

1. The Missing Information Problem

A physician documents an examination in a narrative note.

The biller needs a particular piece of information to complete the claim.

The information may exist in the narrative, but not in the structured field.

The biller searches.

The physician is contacted.

The claim is delayed.

This is not necessarily a coding problem.

It may be an information extraction and workflow problem.

The information was recorded, but the system failed to make it useful at the right moment.

2. The Duplicate Patient Problem

Consider John Smith and Jonathan Smith Jr.

One record lists an abbreviated name.

Another includes a middle name.

The addresses are formatted differently.

The phone numbers may have changed.

Are they the same patient?

Sometimes yes.

Sometimes no.

An overly aggressive matching algorithm can merge two different people.

A poorly designed system can maintain two records for the same person.

Both situations create risk.

AI-powered entity resolution can compare multiple attributes and identify possible matches.

But uncertain matches require appropriate review.

Patient identity is not a guessing game. A mistaken match can become a clinical problem, not merely an administrative inconvenience.

3. The Eligibility Problem

A patient presents an insurance card.

The staff enters the information.

The system shows an insurance company.

Everyone assumes coverage is active.

But the patient may have changed plans.

The policy may have terminated.

The benefits may not cover the intended service.

The practice should verify the appropriate coverage for the relevant date and service.

An insurance card is evidence of an insurance relationship.

It is not proof of current eligibility.

4. The Authorization Problem

A physician orders a service.

The practice assumes authorization is not required.

The appointment is scheduled.

The service is performed.

The claim is submitted.

Then the payer identifies a missing authorization.

Now everyone is looking for the person who was supposed to check.

This is why prior authorization should be treated as a trackable workflow, not a phone call that disappears into someone’s notebook.

A useful system should record:

• Whether authorization is required.
• Which payer rule applies.
• What documentation is needed.
• Who owns the request.
• When it was submitted.
• Whether additional information was requested.
• Whether approval was received.
• Whether the authorization applies to the intended service and date.

An authorization that exists only in someone’s memory is not a reliable authorization-management process.

5. The Denial Problem

A denial arrives.

The biller investigates.

The original encounter occurred weeks earlier.

The person who entered the information may have forgotten the details.

The physician has moved on.

The relevant documentation is buried somewhere.

The biller must reconstruct the event.

This is where the revenue cycle becomes reactive.

The practice is not managing information.

It is investigating history.

And that brings us to one of the most important principles in healthcare administration:

The denial is the crime scene, not necessarily the crime.

The original mistake may have occurred at registration, eligibility verification, documentation, authorization, or coding.

Treating the denial as the beginning of the problem is like treating a fever without asking why the patient has an infection.

Three Expert Perspectives Worth Remembering

These are perspectives drawn from published work and current health IT policy, not personal endorsements of OnnX.

1. Dr. Atul Gawande: Better Systems Beat Heroic Workarounds

Dr. Atul Gawande, surgeon, author, and public-health researcher, has written extensively about the importance of systems, checklists, teamwork, and disciplined execution.

His book The Checklist Manifesto challenges the assumption that expertise alone can overcome complexity.

The lesson applies directly to medical billing.

A practice should not depend on one exceptionally experienced biller remembering every payer’s requirements.

It should build systems that preserve knowledge and make important steps visible.

The best employees should not have to rescue the same broken process every Monday morning.

Takeaway: Do not build a business around people repeatedly compensating for preventable system failures.

2. Dr. Tom Keane: Technology Regulation Should Support Practical Innovation

Dr. Tom Keane, National Coordinator for Health Information Technology, has discussed the need to reduce unnecessary technology development burdens while supporting innovation.

The October 26, 2026 HTI-5 final rule reflects this direction by revising and removing selected requirements in the ONC Health IT Certification Program.

The broader question for independent physicians is whether technology becomes more usable, affordable, and interoperable.

A smaller practice does not need another complicated system that requires additional employees to operate.

It needs technology that reduces work.

Takeaway: Technology should make healthcare operations easier, not create another department to manage.

3. Steven Posnack: Interoperability Must Make Information Useful

Steven Posnack, a health IT policy leader at ONC, has contributed to federal work on health information exchange, interoperability, and standards.

The broader interoperability challenge is not simply transmitting information.

It is ensuring that receiving systems can interpret and use that information appropriately.

A record that arrives without sufficient context can create more work.

A duplicated patient identity can create confusion.

A prior authorization response that cannot be connected to the correct encounter may not solve the administrative problem.

Takeaway: Data exchange is only the beginning. The information must support an actual decision or action.

These perspectives point toward the same conclusion.

Better healthcare technology is not measured by how much information it processes. It is measured by how much unnecessary work it eliminates without compromising accuracy.

The Statistics: Impressive Numbers Do Not Automatically Mean Better Healthcare

Healthcare technology discussions often feature enormous statistics.

Some estimates suggest that healthcare generates approximately 30% of the world’s data. Industry reports also frequently describe rapid growth in healthcare data volumes.

These numbers are interesting.

But here is a more important question:

How much of that information can a physician or biller actually use when a decision needs to be made?

The volume of data does not tell us whether it is accurate.

The number of automated transactions does not tell us whether they were completed correctly.

The number of claims processed does not tell us whether the practice collected the money it was entitled to receive.

For independent practices, I would focus on these measurements:

First-pass claim acceptance: How many claims pass initial submission without rejection?

Denial rate: What percentage of submitted claims are denied?

Days in accounts receivable: How long does it take to collect outstanding payments?

Eligibility accuracy: How often does recorded insurance information match verified coverage?

Authorization completion: How frequently is required authorization obtained before the service?

Manual rework hours: How much time do employees spend correcting avoidable problems?

Exception resolution time: How long does it take to resolve incomplete or conflicting information?

Data accuracy: How often does recorded information match the verified source?

Here is the point.

A software company may proudly report that its system processed one million transactions.

I would rather know how many transactions required no human correction.

Throughput is not productivity when the output is wrong.

Recent News: Healthcare Technology Is Changing. Are Medical Practices Ready?

The timing of this discussion matters.

Healthcare technology policy is evolving, and independent practices should pay attention.

October 26, 2026: HHS Finalizes the HTI-5 Rule

The Department of Health and Human Services announced the HTI-5 final rule, which revises the ONC Health IT Certification Program.

The government projects more than $1.5 billion in regulatory cost savings for health IT developers.

The rule removes 14 of 60 certification criteria and revises seven.

This is an important development because it changes the regulatory environment for health IT developers.

But physicians should ask a more practical question:

Will these changes lead to more affordable, usable, and interoperable technology for independent practices?

That remains to be demonstrated.

Reducing developer compliance burdens does not automatically reduce a clinic’s administrative workload.

October 29, 2026: Healthcare Stakeholders Examine the HTI-5 Implications

The American College of Cardiology published an analysis of the final rule and its implications for health IT certification.

One issue deserves particular attention: changes to requirements concerning the reporting of information about AI-enabled health IT capabilities.

For physicians, the broader lesson is that innovation and transparency must remain in balance.

A practice should understand what an AI system is intended to do, where it may fail, and when human review is required.

Removing regulatory requirements does not remove the need for responsible purchasing decisions.

January 1, 2027: A Major Interoperability Milestone Approaches

CMS established implementation requirements for certain impacted payers under the 2024 Interoperability and Prior Authorization final rule.

Beginning in 2027, applicable payers must implement specified APIs, including Provider Access, Payer-to-Payer, and Prior Authorization APIs, according to their applicable compliance dates.

The goal is to improve access to information and reduce administrative friction.

For medical practices, this creates an opportunity to reconsider how eligibility, authorization, and payer communication are managed.

But do not confuse an API with an operational solution.

An electronic connection can move information.

It cannot guarantee that the information is accurate, complete, or acted upon.

The next competitive advantage will not simply be connecting systems. It will be making connected information useful.

Seven Mistakes Medical Practices Should Stop Making

1. Buying AI before identifying the problem.

A vendor promises to automate everything.

But if the biggest problem is eligibility, purchasing an AI coding product may not help.

Identify the bottleneck first.

2. Assuming AI can fill in every missing field.

AI can locate information and flag missing values.

It should not invent clinical measurements, diagnoses, or treatment details.

3. Treating a complete record as an accurate record.

A database with every field populated can still contain incorrect information.

Completeness and accuracy must be measured separately.

4. Assuming FHIR automatically fixes interoperability.

FHIR provides standards for exchanging information.

It does not eliminate poor mapping, incorrect source data, or incompatible workflows.

5. Automating unnecessary steps.

If a workflow is inefficient, automation may simply make the inefficiency faster.

Redesign the process before automating it.

6. Depending on one employee’s memory.

If only one person knows how to resolve a particular payer problem, the practice has a knowledge-management problem.

Preserve the process.

7. Measuring software activity instead of financial outcomes.

More automated transactions do not necessarily mean more collected revenue.

Measure the results.

A Practical Eight-Step Plan for Independent Practices

You do not need a multimillion-dollar technology project.

Start with the problems you already know exist.

Step 1: Map the revenue cycle.

Document the entire patient journey, from registration to payment.

Identify every handoff.

Ask where information gets lost or repeatedly entered.

Step 2: Find the recurring failures.

Review recent claims and identify the most common reasons for rework.

Do not rely exclusively on employee opinions. Examine actual claim and operational data.

Step 3: Establish a baseline.

Measure denial rates, first-pass acceptance, eligibility errors, accounts receivable aging, and manual rework.

You need a starting point.

Step 4: Introduce targeted automation.

Start with repetitive tasks such as insurance information extraction, eligibility monitoring, document classification, and missing-field detection.

Step 5: Use AI to organize unstructured information.

Apply appropriate natural language processing and document intelligence to identify relevant information in clinical notes and supporting documents.

Preserve the original source.

Step 6: Add validation rules.

Use deterministic rules to check required information, dates, known payer requirements, and coding consistency.

Distinguish between errors and issues requiring review.

Step 7: Create an exception-management process.

Every unresolved problem needs an owner.

The system should show what is missing, who is responsible, and what must happen next.

Step 8: Review the results monthly.

Compare actual performance with the baseline.

If the technology saves time, reduces errors, and improves collections, expand it.

If it creates additional work, reconsider the implementation.

The objective is measurable improvement.

Not another software subscription.

Legal Implications: AI Does Not Become the Responsible Party

A common misconception is that using an AI vendor transfers responsibility for billing accuracy and compliance.

It does not.

Medical practices must consider applicable federal and state requirements.

HIPAA

If a technology vendor creates, receives, maintains, or transmits protected health information on behalf of a covered entity, the vendor may qualify as a business associate.

An appropriate business associate agreement may be required.

Practices should examine:

• Data access.
• Permitted uses.
• Encryption.
• Security controls.
• Audit trails.
• Retention and deletion.
• Breach notification.
• Subcontractor arrangements.

Billing and Documentation Integrity

AI-generated recommendations must not be treated as established clinical facts.

The submitted claim must be supported by the applicable documentation and billing requirements.

Physicians should not approve unsupported diagnoses simply because an algorithm suggests them.

False Claims Exposure

Incorrect claims and unsupported billing can create exposure under applicable federal and state laws.

The specific legal consequences depend on the facts, applicable rules, and circumstances.

AI does not provide immunity.

California Privacy Requirements

California practices should also evaluate applicable state privacy and security requirements, including the California Confidentiality of Medical Information Act where applicable.

Qualified legal and compliance professionals should review higher-risk implementations.

Automation changes how work is performed. It does not eliminate professional responsibility.

Ethical Considerations: Patients Are Not Data Points

There is something dangerous about treating healthcare information as if it were just another business dataset.

A patient is not a spreadsheet row.

A duplicate record is not merely a database inconvenience.

A missing clinical value is not an invitation to estimate.

An incorrect identity match can have consequences far beyond reimbursement.

Responsible healthcare AI requires:

• Accuracy.
• Privacy.
• Transparency.
• Fairness.
• Accountability.
• Human oversight.

The system must be able to distinguish between verified information, extracted information awaiting confirmation, and an algorithmic suggestion.

That distinction should never disappear simply because automation is convenient.

The purpose of AI in healthcare should be to reduce administrative friction, not to introduce a new source of clinical uncertainty.

Myth Busters: Five Things Healthcare Technology Vendors Should Explain

Myth 1: AI eliminates medical billing errors.

Reality: AI can identify patterns and prevent certain errors, but it cannot guarantee that source information is correct or that every payer decision will be favorable.

Myth 2: More data means better decisions.

Reality: More data without context can create more confusion.

Myth 3: FHIR solves all interoperability problems.

Reality: FHIR improves data exchange. Accurate mapping, validation, implementation, and workflow integration are still necessary.

Myth 4: AI can safely repair every incomplete record.

Reality: AI can identify missing information and locate potential sources. It should not manufacture clinical facts.

Myth 5: Medical billing is simply a coding problem.

Reality: Billing is the final stage of a much larger operational process involving registration, eligibility, authorization, documentation, coding, claim submission, and payment.

The fifth myth is perhaps the most expensive.

Because when practices treat billing as an isolated department, they often ignore the upstream problems creating the downstream workload.

The OnnX Perspective: Stop Fixing the Same Problem Twice

This is where my perspective as a physician-entrepreneur and founder of OnnX comes in.

OnnX is built by a physician and a medical biller.

That combination is intentional.

Medical billing is not simply a software problem.

It requires understanding clinical documentation, payer requirements, administrative workflows, coding, reimbursement, and the realities of operating an independent medical practice.

I have seen how much operational knowledge can become concentrated in a few experienced employees.

They know which payer requires what.

They remember which authorization is pending.

They know which claims need attention.

They understand why a particular account has been sitting in accounts receivable for weeks.

But what happens when they leave?

The knowledge leaves with them.

The practice starts asking the same questions again.

Who handled this?

Was authorization obtained?

Did we receive the documentation?

Why was this claim rejected?

What did the payer say last time?

That is not institutional knowledge.

That is institutional memory loss.

A business should not depend on someone remembering what the system should have remembered.

The OnnX philosophy is built around a different way of thinking.

Instead of treating each claim as an isolated transaction, we believe the entire revenue cycle should be understood as a connected operational process.

Patient information affects eligibility.

Eligibility affects authorization.

Authorization affects the encounter.

The encounter creates documentation.

Documentation supports coding.

Coding supports the claim.

The claim produces a payer response.

Payment and denial information should inform the next action.

Every stage is connected.

The technology should understand those connections.

From Data to Workflows to Revenue

Our central philosophy is:

Data → Workflows → Revenue

First, capture and organize information.

Second, connect that information to the correct workflow.

Third, make sure the work reaches an appropriate outcome.

This means moving beyond software that simply displays information or generates recommendations.

A useful system should help answer:

• What information is available?
• What information is missing?
• Which requirement applies?
• What has already been completed?
• What remains unresolved?
• Who owns the next action?
• What happened to the claim?
• What should happen next?

It should preserve the history of important actions.

It should identify exceptions.

It should reduce unnecessary handoffs.

It should make work traceable.

Most importantly, it should help independent practices operate without depending on unnecessary intermediaries.

The goal is not to replace experienced billers with algorithms. It is to stop wasting experienced billers’ time on preventable administrative problems.

That is a very different proposition.

What the Future of Medical Billing Should Look Like

I believe the next generation of medical billing technology will be defined by several developments.

Intelligent Workflow Orchestration

Instead of automating isolated tasks, systems will coordinate connected workflows.

A missing authorization should trigger a defined process.

An incomplete claim should create an actionable exception.

A payer response should update the relevant account.

An unresolved task should remain visible until someone addresses it.

Predictive Compliance Monitoring

Systems will increasingly identify potential documentation and payer-rule problems before claims are submitted.

But predictions must remain distinguishable from verified facts.

Better Interoperability

FHIR and related standards will continue to influence how healthcare information is exchanged.

The opportunity is to connect systems without forcing small practices to manage increasingly complicated administrative infrastructure.

Privacy-Preserving Analytics

Federated learning and other privacy-preserving approaches may allow organizations to develop insights without centralizing all underlying patient information.

However, privacy protection still requires careful technical and organizational controls.

Auditable AI

The next generation of AI should be able to demonstrate:

• What information it accessed.
• What it extracted.
• What it recommended.
• What action it performed.
• What information it changed.
• When human review occurred.
• Whether the intended outcome was achieved.

An AI system should not simply tell us what it did. It should leave enough evidence for us to verify what it did.

That is how trust is built.

Final Thoughts: The Most Expensive Mistake May Be the One Nobody Notices

Let’s return to the patient with the fractured wrist.

The physician did the examination.

The diagnosis was documented.

The treatment was completed.

The patient went home.

But the claim did not get paid on time.

Somewhere in the process, information failed to connect.

Maybe eligibility was not verified.

Maybe the authorization was missing.

Maybe the documentation was incomplete.

Maybe the biller had to reconstruct the encounter.

The practice spent time and money correcting something that could have been prevented.

And tomorrow, another patient will arrive.

Another claim will be submitted.

Another employee will search through another chart.

Unless the underlying process changes, the same problem will happen again.

Here is the contrarian argument:

Healthcare does not need to become better at fixing billing mistakes. It needs to become better at preventing the information failures that create them.

We should stop celebrating how quickly we can process claims and start asking why so many claims require intervention in the first place.

We should stop treating denial management as the primary solution and start examining what happened before the denial.

We should stop building technology around isolated transactions and start building systems around connected information and accountable workflows.

And we should stop confusing artificial intelligence with operational intelligence.

AI is a tool.

Data is the foundation.

Workflows are the machinery.

Revenue is the result.

At OnnX, we believe independent medical practices deserve technology that helps them retain control, reduce unnecessary administrative friction, and operate with greater financial predictability.

Because physicians did not spend years learning medicine to become full-time insurance investigators.

And medical billers should not spend their days searching for information that the system should have preserved.

The future of medical billing will not be won by whoever automates the most steps. It will be won by whoever eliminates the most unnecessary steps.

Get Involved: Help Change the Conversation

I want to hear from physicians, clinic owners, practice administrators, and medical billers.

Here is the question: If you could eliminate one recurring problem in your practice tomorrow, would you choose eligibility verification, prior authorization, documentation gaps, claim denials, or the disconnected systems that create all four?

Share your experience in the comments.

What is the one administrative problem that consumes the most time in your practice without adding value to patient care?

And if this perspective challenges how you think about healthcare technology, share the article with another physician or practice owner.

The conversation should not be limited to software developers and hospital executives.

Independent practices deserve a voice in shaping the future of healthcare technology.

Better information. Better workflows. More predictable revenue.

That is a movement worth building.

If you agree, consider reposting this article to help more independent physicians and clinic owners join the conversation.

PS: A complimentary resource is available in the Featured section of my LinkedIn profile. No registration or email signup is required. Explore it, share your thoughts, and help move the conversation forward.

About the Author

Dr. Daniel Cham, MD, is a physician-entrepreneur, medical consultant, and founder of OnnX, an AI-powered medical billing SaaS platform designed for small and medium-sized medical practices.

His work focuses on the intersection of clinical operations, healthcare technology, medical billing, and practice management.

Drawing on experience in medicine and medical billing, he examines how better information, intelligent automation, and accountable workflows can help independent physicians reduce administrative friction and build more sustainable practices.

His philosophy is simple: healthcare technology should solve problems, preserve institutional knowledge, and give physicians greater control over their practices.

Disclaimer

This article provides general educational and informational perspectives on healthcare technology, clinical data management, artificial intelligence, and medical billing. It does not constitute medical, legal, coding, or regulatory advice. Requirements vary by jurisdiction, payer, organization, and individual circumstances. Consult appropriately qualified professionals before making clinical, legal, compliance, or operational decisions.

Continue the Conversation

Healthcare is changing. The challenges facing independent medical practices are changing with it.

Explore more practical insights, thoughtful discussions, and behind-the-scenes perspectives on healthcare management, medical billing, technology, and innovation.

Continue learning and connect through these platforms:

Connect: LinkedIn | Website | YouTube| Spotify

Knowledge creates progress. Curiosity challenges convention. Better information creates better decisions. Keep learning, keep questioning, and help shape the future of independent healthcare.

References and Further Reading

1. The Denial Isn’t the Problem. It’s the Clue. — Why billing denials often reveal upstream information and operational failures.

2. Your Medical Billing Problem May Not Be a Billing Problem: The Hidden Cost of Creating Work You Should Never Have to Do — How fragmented information creates unnecessary administrative work and financial losses.

3. The Cheapest Claim Is the One You Never Have to Fix — Why upstream intelligence and prevention are more valuable than downstream claim correction.

4. Atul Gawande: A Conversation About Better Healthcare Systems — Systems thinking, diligence, and the importance of improving how healthcare work is performed.

5. HHS Finalizes HTI-5 Deregulatory Rule — The October 2026 federal action to streamline the Health IT Certification Program.

6. HTI-5 Final Rule: What It Means for Health IT — Industry analysis of changes to certification requirements, transparency, and AI-related provisions.

7. ONC HTI-5 Final Rule — Official information on the revised certification framework and FHIR-based interoperability direction.

#OnnX #HealthcareAI #MedicalBilling #RevenueCycleManagement #HealthcareTechnology #MedicalPracticeManagement #IndependentPhysicians #PhysicianEntrepreneur #HealthcareInnovation #ArtificialIntelligence #ClinicalData #DataQuality #FHIR #HealthcareInteroperability #MedicalCoding #ClaimDenials #PriorAuthorization #InsuranceEligibility #PracticeOwners #HIPAA #HealthIT #DigitalHealth #HealthcareAutomation #MedicalPractice #PhysicianLeadership