Healthcare Has a Billing Problem. Or Does It Have an Information Problem?

A provocative look at medical billing, AI automation, payer complexity, and why independent medical practices keep paying for mistakes they didn’t necessarily make.

“They’re able to review things faster, deny things faster, and we’re still pretty manual in terms of our appeals efforts.” — Shawntea “Taya” Gordon, MBA, FACMPE, CEO of Atlas & Perpetua Healthcare Consulting, discussing the growing imbalance between payer automation and provider-side operations,

Here is an uncomfortable thought: What if the medical billing industry is getting very good at fixing the wrong problem?

A physician sees a patient. The examination is completed. The diagnosis is documented. The treatment plan is established. The patient leaves.

Then the real adventure begins.

The insurance information does not match. The authorization is missing. The claim gets rejected. The billing department opens another work queue. Someone calls the payer. Someone else searches the chart. Another person sends an appeal.

Congratulations. The practice has just created another administrative job without hiring anybody.

The physician delivered healthcare.

The billing team delivered detective work.

And somewhere between the examination room and the insurance company’s payment department, the original transaction became an administrative obstacle course.

Sound familiar?

For independent physicians and small medical practices, this is more than an inconvenience. It affects cash flow, employee productivity, patient relationships, and the financial survival of the practice.

But I want to challenge the conventional explanation.

Perhaps healthcare does not have a billing problem. Perhaps it has an information problem that eventually becomes a billing problem.

And if that is true, we have been spending too much time fixing claims and not enough time examining why the information behind those claims keeps breaking.

The uncomfortable truth: Your billing department may be cleaning up somebody else’s mess

Let’s start with a simple example.

A receptionist enters a patient’s date of birth incorrectly.

One digit.

Not a complicated medical decision. Not a sophisticated coding question. Just one digit.

The eligibility response may not match. The claim may fail front-end validation. The billing team investigates. Someone corrects the information. The claim gets resubmitted.

How much did that one digit cost?

Not just the employee’s time.

Consider the chain reaction:

• Registration has to be reviewed.
• Insurance information has to be checked.
• Eligibility may need to be verified again.
• The clearinghouse response has to be interpreted.
• The claim has to be corrected.
• The billing team has to resubmit it.
• Payment may be delayed.

One digit has become a small administrative production.

Now imagine that process repeated across hundreds of encounters.

This is why the cost of an error is often much larger than the cost of correcting the error.

And here is the contrarian part.

We should stop assuming that every billing problem belongs to the billing department.

A claim is the final output of an information chain.

The problem may have started at registration, scheduling, eligibility verification, authorization, clinical documentation, coding, or information exchange between systems.

The billing department may simply be the first place where somebody notices.

The person who discovers the error is not necessarily the person or process that created it.

That distinction matters.

Blaming employees for systemic problems is easy. Designing a workflow that prevents those problems is harder.

But the latter is what actually improves operations.

The great healthcare technology paradox: More software, more administrative work?

Healthcare has invested heavily in electronic health records, practice management systems, clearinghouses, billing platforms, and specialized revenue cycle management services.

Yet physicians continue to complain about administrative complexity.

How is that possible?

We have digitized the paperwork, but have we actually simplified the work?

Consider a typical independent practice.

One system manages scheduling. Another maintains clinical documentation. A third handles claims. A separate portal checks eligibility. Another website manages authorization requests.

The employees become the integration layer.

They copy information, verify information, reconcile information, and occasionally discover that two systems disagree about the same patient.

That is an expensive way to connect software.

It is also a curious definition of automation.

If your employee has to move information between five systems, your practice may have automated the software but not the workflow.

The industry often measures technological progress by the number of functions a product offers.

Physicians should ask a different question:

How many administrative steps can this technology eliminate without creating another problem?

A dashboard with 25 features is not necessarily better than a system that prevents one recurring source of errors.

More technology is not the same as better technology.

And more automation is not necessarily more efficiency.

Sometimes the most sophisticated solution is to stop making people do unnecessary work.

The AI debate: Should we train people better or replace the process with machines?

The original debate presents two competing philosophies.

One side argues that human discipline and structured workflows are the foundation of reliable medical billing.

The other argues that payer complexity has become too extensive for humans to manage without adaptive AI.

Both arguments have merit.

Both also miss something important.

The human accountability argument

The traditional approach begins with a simple principle:

Verify before you bill.

Patient demographics should be accurate. Insurance coverage should be checked. Authorization requirements should be understood. Clinical documentation should support the services provided.

These are foundational responsibilities.

AI should not become an excuse for careless registration or incomplete documentation.

But expecting a front-desk employee to memorize every payer rule, benefit limitation, authorization requirement, and coding update is unrealistic.

A receptionist is not a walking payer-policy database.

And if the practice’s entire revenue cycle depends on one experienced employee remembering everything, that is not a robust system.

That is an operational vulnerability wearing a name badge.

The autonomous AI argument

The technology-centered position makes another valid observation.

Payer requirements change. Documentation standards evolve. Different insurance plans have different rules. Staff turnover creates gaps in institutional knowledge.

AI can help monitor information, identify inconsistencies, and reduce repetitive administrative tasks.

It can flag a missing authorization. It can compare an eligibility response against available patient information. It can identify recurring rejection patterns.

It can also help organize information before the claim reaches the clearinghouse.

But there is a problem with the idea of completely autonomous billing.

AI cannot responsibly invent information simply because a required field is empty.

A missing diagnosis is not an invitation to create one.

A coding inconsistency is not permission to change the clinical record.

A medical necessity denial is not proof that the physician failed to provide appropriate care.

The system must distinguish between what it knows, what it can infer, and what requires human judgment.

The third position: Human-supervised, AI-powered operations

I believe the more useful debate is not human versus AI.

It is whether we can design systems that allow humans and AI to perform the work each is best equipped to handle.

Humans contribute clinical judgment, accountability, contextual understanding, and exception handling.

AI contributes speed, pattern recognition, validation, consistency, and administrative assistance.

Structured information connects the two.

The goal should be:

Capture → Structure → Preserve → Validate → Propagate → Act → Learn.

Instead of waiting for a claim to fail, the system evaluates whether the necessary information is available, consistent, and current before the next step.

That is a fundamentally different way to think about revenue cycle management.

The numbers: Healthcare’s administrative headache is not imaginary

Let’s look at what the latest industry research tells us.

Experian Health’s 2026 State of Claims survey, released October 28, included 204 healthcare professionals involved in finance, billing, registration, reimbursement, claims, and collections.

The findings are revealing.

Claim denial statistics

• 40% reported that at least 10% of their claims were denied.
• 75% said claim denials were increasing.
• 78% said payer policies were changing more frequently.
• 79% said obtaining reimbursement had become more challenging.
• 64% said reimbursement timelines were increasing.
• 89% said reducing denials was a priority in 2026.

The most frequently cited denial reasons included:

• 53%: Services not covered.
• 31%: Frequently changing payer policies.
• 25%: Coding inaccuracy.
• 25%: Difficulty obtaining prior authorization.

These percentages represent the share of respondents identifying reasons among their leading denial causes, not the percentage of all claims denied for each reason.

There is another finding worth considering.

Only 30% of respondents reported recovering between 80% and 100% of denied revenue.

In other words, correcting the claim does not necessarily mean recovering the money.

That is an important distinction for independent practices.

A large health system may have a dedicated denial management department. A small practice may have one or two employees handling billing, collections, insurance calls, and everything else that happens before lunch.

The same denial can have very different operational consequences.

The automation statistics

An October 20 MGMA poll of 214 respondents found that:

• 63% were moving toward more revenue cycle automation.
• 26% expected no major change.
• 11% were moving toward more manual operations.

The direction is clear.

Healthcare organizations want more automation.

But wanting automation and implementing useful automation are two different things.

The uncomfortable conclusion

We are seeing substantial investment in technology while reimbursement complexity remains a major operational challenge.

Perhaps the question should no longer be:

How do we process claims faster?

Perhaps it should be:

Why are we still spending so much time correcting information that should have been reliable before the claim was created?

Recent news: Three developments that deserve a closer look

The healthcare technology conversation is changing quickly. Three October 2026 developments illustrate why.

Development one: Denials are changing, not disappearing

On October 30, Medical Economics reported that services not covered had become the leading denial concern in Experian Health’s 2026 survey.

This is interesting because it challenges the assumption that most billing problems are simply registration mistakes.

Patient demographic accuracy remains important, but payer coverage policies and changing requirements are also major contributors.

A practice can submit a technically clean claim and still encounter a coverage problem.

The lesson: Clean claims are necessary, but they are not the same as payable claims.

Development two: Medical groups are moving toward automation

MGMA reported on October 21 that 63% of surveyed medical groups were moving toward more revenue cycle automation.

Physicians Practice also reported on October 27 that coding, prior authorization, denials, and payment posting were leading areas of interest.

Julia Rosen, senior vice president of IT at MGMA, emphasized the importance of examining existing EHR capabilities before adding more standalone AI tools.

That is sensible advice.

Before buying another platform, ask whether your current systems already provide the functionality you need.

Otherwise, your practice may end up with another subscription and another password.

Development three: Is healthcare AI making the system more expensive?

On October 14, Becker’s Hospital Review reported on a Peterson Health Technology Institute analysis questioning whether administrative AI tools are actually reducing healthcare costs.

The report argued that AI applications involving documentation, coding, billing, and prior authorization could reinforce existing fee-for-service incentives and increase administrative activity.

Caroline Pearson, executive director of the institute, warned:

“On our current trajectory, AI is set to exacerbate existing system failures.”

That deserves attention.

The objective of healthcare AI should not be to make inefficient processes run faster.

If an inefficient workflow takes 20 steps, automating all 20 steps may simply give us a faster inefficient workflow.

The real innovation is eliminating unnecessary steps.

Three expert perspectives: What should physicians take away?

The following perspectives come from healthcare leaders and researchers discussing the changing economics and operations of healthcare technology.

Expert one: Shawntea “Taya” Gordon, MBA, FACMPE

CEO, Atlas & Perpetua Healthcare Consulting

Gordon has highlighted the growing imbalance between automated payer reviews and largely manual provider appeals.

Her observation is particularly relevant to independent practices.

Payers can use technology to examine claims at scale. Provider organizations may still be relying on manual processes to understand and challenge those decisions.

The practical lesson:

Practices need better visibility into denials, faster identification of recurring problems, and a more systematic approach to appeals.

My interpretation: If the payer can process a denial in seconds but the physician’s office needs several days to understand it, the technology imbalance is already part of the business problem.

Expert two: Julia Rosen

Senior Vice President of IT, Medical Group Management Association

Rosen’s advice is to examine the capabilities already available in existing EHR systems before introducing additional standalone AI products.

This is especially important for smaller practices.

A technology purchase should solve a documented problem, not simply respond to an impressive demonstration.

My interpretation: The first question should not be, “Which AI product should I buy?” It should be, “Where does my current workflow break?”

Expert three: Caroline Pearson

Executive Director, Peterson Health Technology Institute

Pearson has challenged the assumption that administrative AI automatically creates healthcare savings.

Her concern is that existing payment incentives can encourage more activity rather than better outcomes.

That is a crucial distinction.

An AI system can generate more documentation, more coding activity, and more automated interactions without reducing the underlying cost of care.

My interpretation: If AI increases the number of administrative transactions but does not reduce the work required to deliver and collect payment for care, what exactly have we improved?

Three experts. Three different concerns. One common lesson.

Gordon highlights the imbalance.

Rosen emphasizes practical integration.

Pearson questions the economics.

Together, they point toward a more disciplined approach to healthcare technology.

Do not automate blindly. Do not buy software impulsively. Measure what actually improves.

The rejection-versus-denial problem: Stop treating two different problems as one

Medical billing has its own vocabulary.

Unfortunately, the vocabulary is not always used consistently.

A claim rejection and a claim denial are not the same thing.

Rejection: The claim did not get through the front door

A rejection generally occurs before payer adjudication.

A clearinghouse or payer front-end system identifies a problem that prevents normal processing.

Examples include:

• Missing information.
• Invalid patient identifiers.
• Incorrect provider identifiers.
• Invalid data formats.
• Inconsistent claim elements.

The 277CA transaction can communicate claim acknowledgment and front-end status information.

Think of it as the claim’s initial checkpoint.

Denial: The payer reviewed the claim and refused payment

A denial occurs after payer adjudication.

Possible reasons include:

• Coverage exclusions.
• Missing authorization.
• Medical necessity requirements.
• Coding problems.
• Timely filing.
• Benefit limitations.
• Coordination-of-benefits issues.

The 835 Electronic Remittance Advice communicates payment and adjustment information.

The distinction matters because the solutions differ.

A demographic error needs correction.

A coverage issue needs plan-specific investigation.

A medical necessity issue requires examination of the clinical documentation and applicable coverage criteria.

A rejection asks whether the claim can proceed. A denial raises the question of whether the payer will pay.

Confusing the two is like taking your car to a mechanic because it failed an emissions test when the actual problem is that it has no engine oil.

Both are problems. They are not the same problem.

Seven practical ways to stop feeding the denial machine

Here is where the discussion becomes useful.

A small practice does not need a million-dollar technology transformation to begin improving its revenue cycle.

It needs a better understanding of where the work is breaking.

1. Stop treating patient registration as data entry

Registration is the beginning of the financial information chain.

Validate:

• Legal name.
• Date of birth.
• Insurance identification.
• Subscriber information.
• Relationship to subscriber.
• Contact information.
• Coordination-of-benefits information.

Do not assume yesterday’s information is still accurate today.

Practical action: Establish a consistent demographic and insurance verification process.

2. Verify eligibility, but understand its limitations

The X12 270/271 transactions support electronic eligibility inquiries and responses.

However, eligibility is not a guarantee of payment.

Coverage, benefits, authorization, network status, and medical necessity can still affect reimbursement.

Practical action: Preserve the eligibility response, verification date, and unresolved coverage questions.

3. Stop managing authorization through memory

An authorization should not exist only in someone’s email or on a sticky note.

Track:

• Whether authorization is required.
• Payer and plan.
• Requested service.
• Submission date.
• Supporting documentation.
• Approval status.
• Authorization number.
• Effective and expiration dates.
• Approved services and limits.

Practical action: Maintain a centralized authorization record connected to the patient and encounter.

4. Do not let AI manufacture clinical documentation

AI can identify missing information and suggest areas requiring review.

It cannot legitimately create a diagnosis simply because the claim needs one.

Practical action: Require documentation-supported coding and human review of material clinical inconsistencies.

5. Read the rejection before resubmitting the claim

Repeatedly resubmitting the same incorrect claim is not a strategy.

It is repetition with optimism.

Practical action: Categorize rejection reasons by payer, provider, service, source, and corrective action.

6. Use denial information as operational intelligence

A recurring authorization denial may reveal a scheduling problem.

A recurring coverage denial may indicate that payer policies have changed.

A documentation-related denial may indicate a communication gap between clinical and administrative workflows.

Practical action: Review patterns, not just individual claims.

7. Measure the cost of administrative friction

Track:

• First-pass acceptance.
• First-pass payment.
• Rejection rate.
• Denial rate.
• Days in accounts receivable.
• Denial recovery.
• Cost per claim.
• Staff touches per claim.
• Authorization turnaround time.
• Repeat denial rate.

One particularly revealing metric is staff touches per claim.

If AI processes more claims but employees still need to intervene just as often, the practice may not have achieved meaningful automation.

A step-by-step roadmap for modernizing your practice

Do not start with the software.

Start with the workflow.

Step 1: Map the actual process.

Follow the patient from registration to final payment.

Identify every point where information is entered, transferred, checked, or corrected.

Step 2: Identify the recurring failures.

Review recent rejections, denials, and delayed payments.

Group them by reason.

Step 3: Find the original source.

Ask where the information originated, who entered it, whether it was verified, and whether it remained accurate.

Step 4: Standardize the information.

Define consistent fields, validation rules, and ownership responsibilities.

Step 5: Automate repetitive work.

Begin with missing-field detection, eligibility review, authorization tracking, claim completeness, and denial categorization.

Step 6: Establish human escalation.

Define which actions AI can perform and which require human approval.

Step 7: Measure the result.

Compare the new workflow against your original baseline.

If the technology creates more work than it eliminates, reconsider the implementation.

The objective is not to demonstrate that the practice uses AI.

It is to demonstrate that the practice operates better because of it.

Pitfalls: Seven ways to make an expensive problem even more expensive

Pitfall one: Automating a broken workflow.

Garbage in, garbage out. AI simply gives the garbage better transportation.

Pitfall two: Blaming the billing department.

The billing team may be discovering errors created elsewhere.

Pitfall three: Assuming eligibility means payment.

Coverage verification does not guarantee reimbursement for a particular service.

Pitfall four: Buying disconnected AI tools.

Another subscription does not automatically improve integration.

Pitfall five: Giving AI unrestricted coding authority.

Clinical documentation must reflect actual care, not an algorithm’s preferred reimbursement outcome.

Pitfall six: Ignoring vendor dependency.

A consolidated platform may simplify operations, but continuity planning remains essential.

Pitfall seven: Measuring software activity instead of business outcomes.

A thousand automated actions mean very little if the practice’s cash flow and administrative burden remain unchanged.

Myth busters: What the industry needs to stop believing

Myth: AI will eliminate medical billers.

Reality: AI can reduce repetitive work, but human expertise remains essential for exceptions, compliance, and complex payer interactions.

Myth: Clean claims always get paid.

Reality: A technically valid claim can still fail because of coverage, authorization, medical necessity, or contractual requirements.

Myth: Every denial is a billing error.

Reality: The problem may originate in registration, scheduling, clinical documentation, payer policy, or system integration.

Myth: More AI means greater efficiency.

Reality: Poorly designed automation can accelerate existing inefficiencies.

Myth: AI can independently establish medical necessity.

Reality: AI can assist with information review, but clinical judgment and applicable coverage criteria remain essential.

Myth: One integrated platform eliminates all risk.

Reality: Consolidation can reduce integration friction while increasing vendor dependency.

Myth: A denial is simply lost revenue.

Reality: It can also be a signal of a recurring operational problem.

Legal and regulatory implications: AI does not get a compliance exemption

Healthcare technology operates within a complicated legal environment.

A practice cannot outsource its responsibility simply because an algorithm performed the task.

HIPAA and patient information

The HIPAA Privacy, Security, and Breach Notification Rules establish obligations involving protected health information.

Practices should evaluate:

• Business associate agreements where required.
• Access controls.
• Encryption.
• Audit trails.
• Data retention.
• Incident response.
• Subcontractor access.
• Secure information exchange.

Do not assume that every consumer AI platform is appropriate for handling PHI.

Billing and coding integrity

Claims must be supported by accurate documentation and applicable coding rules.

AI suggestions should not become unsupported billing entries.

The physician’s record must accurately represent the care provided.

False Claims Act

Knowingly submitting false or fraudulent claims to federal healthcare programs can create serious liability under the federal False Claims Act.

AI does not eliminate the responsibilities of the provider or organization submitting the claim.

CMS interoperability and prior authorization

CMS-0057-F, the 2024 Interoperability and Prior Authorization Final Rule, established requirements for specified payers.

Many operational provisions began in 2026, while major API implementation requirements generally begin in 2027, subject to the rule’s specific timelines and exceptions.

The rule is important because it moves the industry toward more standardized information exchange.

But interoperability is not merely about moving data.

The data must retain its meaning, accuracy, and context.

Practices should also monitor applicable state laws, payer contracts, Medicare requirements, and other regulatory developments.

Obtain qualified legal and compliance advice before deploying systems that influence clinical documentation, coding, reimbursement, or coverage decisions.

Ethical considerations: Getting paid is important. Getting it right is more important.

There is a temptation to measure billing technology by one number: reimbursement.

That is insufficient.

The system must also protect clinical integrity.

An AI system should not invent a diagnosis to improve payment.

It should not alter the medical record without appropriate authorization and documentation.

It should not encourage unnecessary services or unsupported coding.

Patients should not experience delays because disconnected systems failed to communicate.

And physicians should know when a machine is making a recommendation, what information supports it, and who is responsible for approving an action.

The ethical objective is simple:

Accurate care. Accurate documentation. Appropriate reimbursement.

Anything else risks turning automation into another source of healthcare dysfunction.

Tools, metrics, and resources: What should a small practice actually use?

A small practice does not need every available technology.

It needs a coherent operational environment.

Consider these categories:

Electronic health records: Clinical documentation, encounters, orders, and patient information.

Practice management systems: Scheduling, registration, billing, and operational workflows.

Eligibility tools: Electronic coverage inquiries and payer responses.

Clearinghouses: Claim submission, acknowledgments, and front-end validation.

ERA processing: Payment and adjustment information from 835 transactions.

Authorization management: Requirements, submissions, approvals, expiration dates, and supporting records.

Analytics: Rejection trends, denial patterns, accounts receivable, and operational performance.

AI-assisted validation: Missing information, inconsistencies, repetitive patterns, and workflow recommendations.

Five questions to ask a technology vendor

1. Does the product integrate with our existing systems?
2. Can it explain why it flagged a claim or patient record?
3. Does it preserve the original source information?
4. Can we audit AI-generated recommendations and human approvals?
5. Can you demonstrate measurable reductions in administrative work?

And one more:

If we stop using your product tomorrow, can we retrieve our data and continue operating?

That question may be more important than the product demonstration.

The OnnX perspective: Stop treating the claim as the starting point

This is where the discussion becomes particularly important to me as a physician and entrepreneur.

At OnnX, the thesis is that healthcare billing is a data quality and context problem, not simply a tooling problem.

The industry often focuses on the claim because that is where the financial consequences become visible.

But the claim is not where the story begins.

The information begins much earlier.

A patient has a demographic record.

That patient has insurance coverage.

The coverage has eligibility conditions, benefits, limitations, and possibly authorization requirements.

The patient has an encounter.

The encounter has clinical documentation.

The documentation supports coding.

The coding contributes to the claim.

The claim enters the payer’s adjudication process.

These are not unrelated pieces of information.

They are connected parts of one operational story.

Yet many systems treat them as separate transactions.

That is the opportunity.

What OnnX is exploring

OnnX is an AI-powered medical billing SaaS concept for small and medium-sized physician-owned practices.

The goal is to reduce administrative friction and help practices manage their revenue cycle with greater visibility and control.

The philosophy is human-supervised and AI-powered.

Not AI replacing physicians.

Not AI inventing clinical justification.

Not another software layer that requires employees to reconcile five different dashboards.

The focus is on structuring and validating information before it becomes a downstream problem.

The operating concept:

Capture: Collect relevant patient, coverage, and encounter information.

Structure: Organize it consistently so it can be interpreted and used.

Preserve: Maintain the context as information moves through the workflow.

Validate: Identify missing, inconsistent, or outdated information.

Propagate: Make relevant information available to the appropriate connected workflows.

Act: Support appropriate administrative actions and escalate uncertain decisions.

Learn: Identify recurring patterns and improve the process.

Consider prior authorization

A traditional reactive process discovers that authorization is missing after the claim has been submitted or denied.

A more proactive process identifies the requirement before the service, tracks the request, records the approval, and connects the evidence to the encounter.

The first process manages a failure.

The second attempts to prevent one.

That is the difference between reactive revenue cycle management and preventive revenue cycle management.

Why independent practices?

Independent practices have little room for unnecessary administrative complexity.

They often operate with smaller teams, tighter margins, limited IT resources, and less capacity to absorb reimbursement delays.

They need technology that is practical and economically defensible.

They do not need another dashboard simply because the dashboard has artificial intelligence in its product description.

The objective is to reduce unnecessary work, improve information continuity, and help physicians preserve their independence.

The physician should be thinking about the patient, not wondering which software system lost the authorization.

Future outlook: The next competitive advantage may be what you prevent

Healthcare AI will continue to evolve.

But I believe several developments will determine which technologies create lasting value.

More intelligent eligibility and authorization

Payer interoperability initiatives may make relevant information more accessible through standardized APIs.

The challenge will be integrating that information into actual practice workflows.

AI-assisted denial prevention

Systems will increasingly analyze rejection and remittance patterns.

The most useful platforms will not simply tell practices what went wrong.

They will help identify where the problem originated and how to prevent recurrence.

Structured information and context preservation

Moving data between systems is not enough.

Systems need to preserve the relationships between patient, coverage, authorization, encounter, documentation, and claim.

Human-supervised AI agents

AI agents will likely take on more repetitive administrative activities.

Their authority should remain bounded by defined rules, audit trails, and escalation requirements.

Greater regulatory scrutiny

Payers, providers, regulators, and patients will increasingly examine how AI affects clinical judgment, coding, medical necessity, and reimbursement.

Measurable return on investment

Healthcare practices will become less impressed by AI demonstrations and more interested in measurable results.

The important questions will be:

• How many errors were prevented?
• How many staff hours were saved?
• How much unnecessary rework disappeared?
• Did first-pass payment improve?
• Did days in accounts receivable decrease?
• Was clinical accuracy preserved?

The future will not belong to the company with the most AI features.

It will belong to the companies that solve meaningful problems without creating new ones.

Frequently asked questions

1. What is the biggest problem in medical billing?

There is no single cause. Problems can originate in inaccurate registration, coverage limitations, authorization, coding, documentation, payer policy, and disconnected systems. Many are information and workflow problems rather than isolated billing mistakes.

2. Can AI eliminate claim denials?

No. AI can help prevent avoidable errors, identify patterns, and improve workflows, but it cannot eliminate payer coverage limitations, contractual disputes, or every documentation and medical necessity issue.

3. What is the difference between a rejection and a denial?

A rejection generally occurs before payer adjudication because the claim fails front-end processing. A denial occurs after the payer adjudicates the claim and determines that payment will not be made or will be reduced.

4. Can AI replace medical billers?

AI can automate repetitive tasks, but human expertise remains important for exceptions, appeals, compliance, and complex payer interactions.

5. Does eligibility verification guarantee payment?

No. It provides information about coverage and benefits. It does not guarantee that a particular service is covered, authorized, medically necessary, or payable.

6. Why should practices focus on upstream data quality?

Because inaccurate or incomplete information introduced early can create multiple downstream problems, increasing administrative work and delaying reimbursement.

7. What should a practice measure when adopting AI?

Rejection rate, denial rate, first-pass payment, days in accounts receivable, denial recovery, cost per claim, and staff touches per claim are useful measures.

8. Is one integrated billing platform always better?

No. Integration can reduce information-transfer problems, but practices must also consider vendor dependency, security, reliability, portability, and contingency planning.

9. Can AI change clinical codes to obtain payment?

AI should not create unsupported diagnoses or modify documentation to secure reimbursement. Coding must accurately reflect the services provided and applicable rules, with appropriate human review.

10. What is preventive revenue cycle management?

It is an approach that identifies and addresses information gaps before they become claim rejections, denials, delayed payments, or costly administrative corrections.

11. What is the most important question to ask an AI billing vendor?

Ask what measurable operational problem the product eliminates and how the vendor proves that the improvement is real.

12. What is OnnX trying to accomplish?

OnnX is exploring AI-powered medical billing technology that helps independent physician practices structure, validate, and manage billing-related information with human supervision and greater workflow continuity.

Final thoughts: Stop celebrating the repair. Start questioning the failure.

There is something strange about an industry that spends so much money correcting information after it has already failed.

We have built sophisticated systems to process claims, analyze denials, automate appeals, and move administrative work faster.

But speed is not the same as progress.

If a practice repeatedly makes the same registration error, faster denial management is not the answer.

If authorization information disappears between systems, another dashboard is not the answer.

If clinical documentation and billing data become disconnected, an AI agent that guesses what is missing is certainly not the answer.

The best denial is the one that never happens.

The best administrative task is the one the physician’s office no longer has to perform.

And the best AI is not necessarily the AI that does the most work. It may be the AI that makes unnecessary work disappear.

Healthcare needs to move beyond the question of how to process claims more efficiently.

It needs to ask how to create better information, preserve context, and prevent avoidable failures before they reach the billing department.

That is a much more interesting problem.

And perhaps a much more important one.

Get involved: Help change the conversation

Let me ask a question that may be uncomfortable for the healthcare technology industry:

Are we actually solving medical billing problems, or are we becoming increasingly sophisticated at managing the consequences of bad information?

I would like to hear from physicians, independent clinic owners, medical billers, and practice administrators.

What is the one administrative problem in your practice that you believe should have been eliminated years ago?

Share your experience in the comments.

If this perspective challenges how you think about medical billing, consider reposting it. Someone in your professional network may be dealing with exactly the same problem.

The future of healthcare technology should be shaped by the people who understand the work.

Question the process. Challenge the assumptions. Help build something better.

About the Author

Dr. Daniel Cham is a physician-entrepreneur and medical consultant with experience in healthcare management, medical practice operations, and medical billing.

He is the founder of OnnX, an emerging AI-powered medical billing SaaS initiative focused on helping small and medium-sized physician-owned practices reduce administrative friction through structured information, intelligent validation, and human-supervised automation.

His interests include healthcare technology, artificial intelligence, revenue cycle management, and the future of independent medicine.

Disclaimer

This article is intended for general educational and informational purposes only. It should not be interpreted as medical, legal, financial, coding, or regulatory advice. Requirements may vary by payer, jurisdiction, and individual circumstances. Readers should consult appropriately qualified professionals before making operational, clinical, or compliance decisions.

Continue the Conversation

Healthcare is changing. The important question is whether our systems are changing in ways that actually help the people delivering care.

Explore more perspectives, practical ideas, and discussions about healthcare operations, medical technology, and innovation.

Connect: LinkedIn | Website | YouTube

Knowledge drives progress. Keep questioning conventional thinking, explore new possibilities, and contribute to a better future for healthcare.

A free resource for your professional development

P.S. Visit the Featured section of my LinkedIn profile for a free resource. No registration and no email signup required.

If you found this article useful, consider reposting it to help other physicians and clinic owners rethink the relationship between medical billing, information quality, and administrative complexity.

References and Further Reading

Industry Research, Reports, and Expert Perspectives

1. Shryock, T. (October 30, 2026). Why are claims being denied in 2026? ‘Services not covered’ now tops the list. Medical Economics.
   Examines the 2026 State of Claims survey, including denial trends, payer policy changes, reimbursement challenges, and the growing gap between denied claims and recovered revenue.
2. Experian Health. (October 28, 2026). Healthcare claim denial statistics: State of Claims Report 2026.
   Provides the underlying survey findings on denial rates, common denial reasons, payer reimbursement challenges, and provider interest in AI and automation. The survey included 204 healthcare professionals involved in financial, billing, registration, reimbursement, claims, and collections decisions.
3. MGMA Revenue Cycle Insights. (October 21, 2026). Most medical groups adding RCM automation amid a wave of payer-driven demands. Medical Group Management Association.
   Reports that 63% of surveyed medical group leaders were leaning toward more revenue-cycle automation in 2026. Also examines coding, prior authorization, denials, appeals, and payment posting as automation priorities.
4. The Hidden Cost of Medical Billing: Why Clinics Are Losing Revenue in Silence (and How AI Is Reshaping the Middle Layer of Healthcare)
   Explores how fragmented workflows, documentation ambiguity, and downstream corrections create administrative costs and revenue leakage.
5. The World Cup Has Better Real-Time Visibility Than Your Medical Billing System — And That’s the Problem
   Examines visibility latency, delayed feedback, and why practices often discover revenue-cycle problems too late.

Hashtags

#MedicalBilling #RevenueCycleManagement #HealthcareAI #ArtificialIntelligence #HealthcareTechnology #PhysicianEntrepreneur #IndependentPractice #MedicalPracticeManagement #ClaimDenials #PriorAuthorization #HealthcareInnovation #DigitalHealth #PhysicianLeadership #OnnX