One nurse's journey through misread signals, severe aortic stenosis and an unexpected pancreatic tumor exposes a deeper problem in modern healthcare: we don't necessarily lack information—we lack the attention, coordination and workflow needed to act on what matters.
“Use AI as a tool, not a replacement.” — Zach
Bush, MD
There is something deeply unsettling about a nurse becoming
the patient.
A nurse knows the language.
She knows the questions.
She knows what normal looks like.
She knows that a symptom can be more important than the
diagnosis currently sitting at the top of the chart.
And sometimes, perhaps most importantly, she knows when
something doesn't feel right.
Lois Bockmann knew.
Bockmann spent much of her career caring for other people.
She worked as an operating-room nurse and later served as a lead nurse at
WakeMed North Surgery Center.
Then the bed was hers.
Bockmann was born with a congenital bicuspid aortic valve
and had been monitored for her heart condition for years. She developed an
arrhythmia in 2015.
But beginning in September 2025, she noticed something
different.
Her shortness of breath was getting worse.
Her heart rhythm was becoming increasingly problematic.
She sought medical attention.
According to WakeMed's account of her experience, the care
she received elsewhere remained heavily focused on the arrhythmia while her
concerns about worsening shortness of breath were not adequately addressed.
Months went by.
September.
October.
November.
December.
January.
By February 2026, Bockmann knew she needed another path.
Her husband had previously undergone coronary bypass surgery
at WakeMed Heart Center with Trevor Upham, MD, FACS.
So they contacted WakeMed.
Bockmann connected with Bryon Boulton, MD, FACS.
An echocardiogram on February 9 finally revealed the
problem:
Severe aortic stenosis.
Boulton recommended a minimally invasive aortic valve
replacement.
But then something unexpected happened.
During the preoperative evaluation, a CT scan revealed a
mass on Bockmann's pancreas.
Suddenly, the patient who had arrived because she couldn't
breathe properly was dealing with two potentially life-changing problems.
The cardiac team didn't simply hand the pancreatic problem
to somebody else and move on.
Boulton contacted Joshua Herb, MD, MSCR, a surgical
oncologist.
Herb evaluated Bockmann while she was still hospitalized.
Further imaging, endoscopy and biopsy followed.
The diagnosis was a pancreatic neuroendocrine tumor.
It was malignant.
But it was also slow-growing and, importantly, surgically
removable.
Bockmann underwent her minimally invasive aortic valve
replacement on March 10.
Then she went through cardiac rehabilitation to become
strong enough for another operation.
On June 15, she underwent a robotic distal pancreatectomy
and splenectomy.
Her care involved a much larger team, including Bob
O'Brien, nurse navigator Kristin Kleber, physician assistants Allie
Gordon and Sarah Fernandez, along with Boulton, Herb and other
specialists.
Two major diagnoses.
Two major surgeries.
Multiple specialties.
One patient.
And, perhaps most importantly, a connected story.
WakeMed's account describes Bockmann's care as coordinated
from the first call through postoperative follow-up. Her clinicians
communicated with one another. Her nurse navigator stayed involved. Her
surgical teams remained connected.
Bockmann later described the experience as fundamentally
different from what she had experienced at other large healthcare facilities.
That is where her story becomes much bigger than one
patient.
Because here's the uncomfortable question:
What if healthcare's biggest problem isn't that we don't
have enough information?
What if we have too much?
And what if the real problem is knowing what deserves
attention now?
The Healthcare Industry Has a Data Problem. And It
Doesn't.
Healthcare has spent decades becoming extraordinarily good
at collecting information.
Electronic health records.
Laboratory results.
Imaging.
Claims.
Prior authorizations.
Medication histories.
Clinical notes.
Messages.
Referral records.
Dashboards.
Alerts.
Work queues.
Analytics.
Artificial intelligence.
We have built an enormous information machine.
And then we gave physicians inboxes.
Lots of inboxes.
That might be the most healthcare sentence ever written.
We built technology to help clinicians manage complexity.
Then we created more complexity for clinicians to manage.
The contradiction is almost funny.
Almost.
Because behind every notification is a person.
Behind every denied claim is a patient's treatment.
Behind every missing document is somebody's time.
Behind every unanswered message is somebody waiting.
Behind every administrative task is somebody who eventually
has to do the task.
And that somebody is often a highly trained clinician or an
already-overloaded staff member.
So perhaps the question isn't:
“How do we give healthcare more information?”
Perhaps it is:
“How do we make healthcare pay attention to the right
information?”
That is a very different problem.
Lois Bockmann's Story Is Really a Story About Attention
Look at what happened to Bockmann.
The arrhythmia was real.
But it wasn't the whole story.
The shortness of breath was real.
But the cause wasn't immediately clear.
Eventually, the right evaluation identified severe aortic
stenosis.
Then another investigation uncovered an entirely different
problem.
The pancreatic tumor was not necessarily sitting there
waiting for somebody to discover it on a giant dashboard.
It was discovered because somebody followed a clinical
workflow far enough to look.
This is important.
Healthcare doesn't fail only when it lacks information.
It can fail when information exists but doesn't trigger the
right action.
That distinction is enormous.
A chart can contain the answer.
A payer portal can contain the denial reason.
A claim can contain the missing field.
A referral can contain the clue.
A prior authorization can contain the documentation
requirement.
A physician can know what needs to happen.
And yet nothing happens.
Why?
Because information isn't action.
Information is potential.
Workflow turns potential into action.
The Contrarian Take: Healthcare May Not Need More AI
I know.
That's an uncomfortable sentence for somebody building an AI
healthcare company.
But it needs to be said.
Healthcare does not necessarily need more AI.
Healthcare needs better workflows.
AI can help.
But AI sitting on top of a broken workflow can simply make
the broken workflow faster.
That isn't innovation.
That's accelerated chaos.
Imagine taking a terrible administrative process and adding
an extremely efficient AI system to it.
Congratulations.
You now have a very efficient terrible administrative
process.
Recent healthcare reporting has already begun documenting
this problem. Medical Economics highlighted concerns that AI layered onto
fragmented administrative processes can actually increase iterations, denials
and appeals rather than reduce them. One analysis described the emerging possibility
of “bot wars,” where providers and payers automate both sides of an already
dysfunctional process.
That should make everyone pause.
Because the industry has a strange habit:
First we automate.
Then we discover we automated the wrong thing.
Then we build another system to manage the automation.
Then we hire someone to manage the system that manages the
automation.
Eventually someone creates a dashboard to tell us why the
original process became more complicated.
At some point, we should probably stop.
The AI Paradox
AI is becoming mainstream in medicine.
The American Medical Association's 2026 research found that 81%
of physicians report using AI professionally, more than double the 38%
reported in 2023. More than three-quarters of physicians also said AI improves
their ability to care for patients.
That's significant.
But adoption isn't the same thing as transformation.
A physician can use AI and still have a terrible workflow.
A practice can purchase six AI tools and still drown in
administrative work.
A hospital can deploy an ambient scribe and still have a
prior authorization nightmare.
A billing department can automate claim submission and still
spend hours chasing denials.
AI adoption is not the finish line.
Workflow improvement is.
That's the distinction I believe healthcare needs to make.
The Best AI May Be the AI You Barely Notice
There is a temptation to make AI visible.
Big dashboards.
Chat interfaces.
Generative summaries.
Animated assistants.
“Copilots.”
“Agents.”
“Autonomous intelligence.”
The vocabulary is impressive.
The workflow can still be miserable.
The best AI in healthcare may actually be almost invisible.
It might quietly recognize that a claim is likely to be
denied.
It might notice that documentation doesn't support a
particular code.
It might retrieve the relevant information.
It might identify the likely problem.
It might prepare the correction.
It might route the task to the right person.
And then it might get out of the way.
No fireworks.
No robot voice.
No dramatic AI avatar announcing:
“Good morning, Dr. Cham. I have optimized your revenue
cycle.”
Please don't.
The physician doesn't need another coworker who talks too
much.
The physician needs fewer things to do.
The Real Currency of Healthcare Is Attention
We usually talk about healthcare resources in terms of
money.
Beds.
Clinicians.
Equipment.
Drugs.
Facilities.
Technology.
But there is another resource that is harder to measure:
attention.
A physician has a finite amount of cognitive bandwidth.
A nurse has a finite amount of attention.
A practice administrator has a finite amount of attention.
A biller has a finite amount of attention.
A patient has a finite amount of attention.
And administrative systems compete for it.
Every interruption has a cost.
Every unnecessary handoff has a cost.
Every duplicate data entry has a cost.
Every portal message that could have been resolved
automatically has a cost.
Every claim that requires five manual touches instead of one
has a cost.
Not always a financial cost.
A cognitive cost.
A human cost.
And eventually, a clinical cost.
Burnout Is Not Just About Working Too Much
Healthcare often discusses burnout as though the solution is
better resilience.
Meditate.
Take vacation.
Exercise.
Practice mindfulness.
Set boundaries.
Those things can help.
But here's the contrarian question:
What if the physician isn't failing to cope with the
workflow?
What if the workflow is failing the physician?
The distinction matters.
You can teach someone to tolerate an inefficient process.
That doesn't make the process efficient.
You can teach a physician resilience.
That doesn't make prior authorization less ridiculous.
You can tell a practice administrator to “work smarter.”
That doesn't eliminate five payer portals.
And you can tell a biller to be more productive.
That doesn't explain why the same denial keeps returning.
The AMA's research shows physicians increasingly view AI as
a way to reduce administrative workload, with 73% of surveyed physicians saying
they expect AI to reduce administrative work through automation.
The opportunity is obvious.
But so is the warning.
Don't automate the burden. Remove it.
Medical Billing Is a Perfect Example
Consider the medical billing workflow.
Patient.
Documentation.
Coding.
Claim.
Payer.
Denial.
Appeal.
Payment.
A/R.
On paper, it looks orderly.
In practice, it can look like a relay race in which everyone
keeps dropping the baton.
A claim is submitted.
It is denied.
Someone opens the denial.
Someone checks the payer portal.
Someone reviews the chart.
Someone searches for documentation.
Someone identifies the probable cause.
Someone contacts the clinical team.
Someone waits.
Someone corrects the claim.
Someone resubmits.
Then the payer finds something else.
Repeat.
The industry often calls this revenue cycle management.
Sometimes it feels more like revenue cycle archaeology.
Someone is digging through three systems looking for
evidence of what happened six weeks ago.
That is not a technology problem alone.
It is a workflow problem.
What If a Denial Were Treated Like a Workflow Event?
Imagine this instead.
A claim is rejected.
The system immediately analyzes the rejection.
It identifies the likely reason.
It compares the denial against the claim.
It retrieves relevant documentation.
It checks whether the documentation supports the correction.
It identifies what is missing.
It recommends the next action.
It prepares the work.
A human reviews it.
The human approves.
The claim is corrected.
The system learns from the outcome.
That's fundamentally different from:
“Here is another dashboard. Good luck.”
The difference is not merely AI.
The difference is workflow design.
This Is Where OnnX Comes In
I founded OnnX around a simple question:
Can medical billing work more like an intelligent
workflow and less like a collection of disconnected tasks?
Not:
“How do we replace everyone?”
Not:
“How do we put AI on every screen?”
And certainly not:
“How do we make a chatbot for billing?”
The better question is:
What work should disappear?
Then:
What work should be simplified?
Then:
What work can AI safely perform?
And finally:
Where must a human remain accountable?
That last question is critical.
Because healthcare isn't a place where “move fast and break
things” is a particularly attractive philosophy.
Breaking a spreadsheet is annoying.
Breaking a patient's care pathway is something else
entirely.
The Human Should Not Be the Workflow's Error-Handling
Mechanism
This is one of the most important ideas in healthcare
automation.
Many organizations design workflows like this:
- Create
automated process.
- Wait
for something to go wrong.
- Send
the exception to a human.
- Human
figures everything out.
That's not intelligent automation.
That's human-powered error recovery.
A better workflow asks:
What can we predict?
What can we validate?
What can we catch earlier?
What information can we assemble automatically?
What decisions can be recommended?
What should be escalated?
What actually requires judgment?
This is the difference between automation and workflow
intelligence.
Expert Opinion Round-Up
Atul Gawande: Better Systems Matter
Atul Gawande has spent years examining the uncomfortable gap
between medical knowledge and reliable execution.
His work repeatedly makes the same broader point: knowing
what to do is not enough.
Healthcare needs systems that help people consistently do
what works.
That idea applies directly to AI.
AI should not merely increase the amount of information
available to physicians.
It should help healthcare organizations reliably execute
better processes.
Knowledge without execution is potential.
Knowledge embedded in a good workflow becomes care.
John Whyte, MD, MPH: AI Should Enhance Physicians
The AMA's 2026 research shows that physicians are
increasingly using AI, but the organization's message remains cautious: AI
should enhance rather than replace physicians and must be safe, effective and
responsibly integrated.
That distinction matters.
The future isn't necessarily physician versus AI.
It may be:
physician plus better workflow.
That's a much more useful conversation.
Rebecca Mishuris, MD, MPH: Technology Can Return Time
Medical Economics recently highlighted the experience of
Rebecca Mishuris, MD, MPH, chief medical information officer at Mass General
Brigham, where burnout reportedly fell substantially following introduction of
ambient documentation technology.
The lesson isn't that every AI product will produce the same
result.
The lesson is simpler:
Time returned to clinicians has real value.
Technology becomes meaningful when the person using it
actually experiences the difference.
The Industry's Favorite Question Is the Wrong Question
Healthcare executives often ask:
“What can AI do?”
I think the better question is:
“What shouldn't humans have to do?”
That shift changes everything.
AI can summarize.
AI can classify.
AI can extract.
AI can compare.
AI can predict.
AI can draft.
AI can route.
AI can monitor.
AI can recommend.
But the most valuable capability may be something much less
glamorous:
AI can remove work.
And removing work is often more valuable than generating
content.
The AI Arms Race Could Make Healthcare Worse
Here's another uncomfortable prediction.
Healthcare organizations may increasingly compete over who
has the most AI.
That could become a mistake.
Imagine two practices.
Practice A has twelve AI tools.
Practice B has three.
Practice A's systems don't communicate.
Practice B redesigned its workflows first and then applied
AI selectively.
Which one is more advanced?
I would bet on Practice B.
Technology count is not a measure of transformation.
Friction eliminated is.
A Better AI Scorecard
Stop asking only:
“How accurate is the model?”
Also ask:
- How
many clicks disappeared?
- How
many handoffs disappeared?
- How
many manual touches disappeared?
- How
much rework disappeared?
- How
quickly does an exception reach the right person?
- How
often does the system escalate appropriately?
- How
often does a human need to correct the AI?
- Does
the physician get time back?
- Does
the staff get time back?
- Does
the patient experience improve?
For revenue-cycle workflows, useful metrics include:
Clean claim rate
Denial rate
Days in A/R
Denial resolution time
Appeal success rate
Manual touches per claim
Rework rate
First-pass resolution
Staff hours spent on denials
And one metric I believe deserves much more attention:
Human touches eliminated per successful workflow.
That is a much more interesting AI metric than the number of
prompts generated.
The Myth: “Automation Means No Humans”
No.
That's the wrong model.
The better model is:
AI handles repetition.
AI handles pattern recognition.
AI prepares the work.
AI surfaces exceptions.
Humans handle judgment.
Humans handle accountability.
Humans handle the situations where context matters more
than pattern.
Human review isn't proof that automation failed.
Sometimes human review is exactly what responsible
automation looks like.
Another Myth: “More AI Means More Efficiency”
Absolutely not.
You can automate a bad workflow.
You can automate redundant work.
You can automate unnecessary approvals.
You can automate the wrong data.
You can automate communication between two systems that
shouldn't be communicating in the first place.
And then you can proudly report:
“We automated 87% of the process.”
Congratulations.
You may have just automated 87% of the wrong process.
This is why workflow mapping should come before AI
deployment.
The 30-Day Workflow Experiment
You don't need a $2 million transformation program to start.
Try this for 30 days.
Days 1–7: Watch the Work
Don't redesign anything.
Observe.
Where do staff members spend time?
Where do physicians get interrupted?
Where do claims stall?
Where do people copy and paste information?
Where do people switch between systems?
Where does someone say:
“I have to check another system.”
Write those moments down.
They're clues.
Days 8–14: Count the Handoffs
For one workflow, count every handoff.
Who starts it?
Who touches it next?
Who reviews it?
Who approves it?
Who sends it?
Who waits?
Who checks it again?
You may discover that the problem isn't the task.
It's the number of people required to move the task.
Days 15–21: Separate Judgment From Repetition
Ask:
What actually requires expertise?
What requires judgment?
What requires authorization?
What is simply retrieval?
What is simply data entry?
What is simply checking?
What is simply routing?
That last category is where automation often becomes
interesting.
Days 22–30: Test One Workflow
Don't automate everything.
Pick one painful workflow.
Measure it before.
Change it.
Measure it afterward.
If nothing improves, don't defend the technology.
Change the workflow.
Or remove the automation.
That's not failure.
That's product development.
What Healthcare Gets Wrong About “Best Practices”
Healthcare loves best practices.
Guidelines.
Protocols.
Checklists.
Standards.
They matter.
But there is a danger.
A best practice from a large academic medical center may not
be the best practice for a five-physician independent clinic.
A workflow designed for a health system with hundreds of IT
employees may be absurd for a small practice.
A solution that requires twelve integrations may technically
be impressive and operationally useless.
The question shouldn't be:
“Is this a best practice?”
It should be:
“Is this the best workflow for this environment?”
Context matters.
Small Practices Deserve Better Technology, Not More
Technology
Independent practices are particularly interesting.
They often don't have armies of analysts.
They don't have endless implementation budgets.
They don't have a dedicated team for every payer.
They can't afford technology that requires a technology
department to operate.
That means simplicity isn't a luxury.
It's a requirement.
The best system for a small clinic may be the one that
quietly removes ten repetitive tasks without requiring the practice to hire
three people to manage it.
This is one reason I believe healthcare AI should become
increasingly workflow-native.
Not another destination.
Not another login.
Not another dashboard.
Something that fits into the work already happening.
The Hidden Cost of Administrative Fragmentation
Let's return to Lois Bockmann.
Her story illustrates the upside of coordination.
Different clinicians.
Different specialties.
Different responsibilities.
One patient.
Communication connected the pieces.
Now imagine the opposite.
The cardiologist doesn't know what oncology discovered.
Oncology doesn't know what cardiology is planning.
The nurse navigator has to reconstruct the story.
The patient becomes the messenger.
The patient carries the information from one office to
another.
That happens in healthcare.
And it is exhausting.
We sometimes call it fragmentation.
I would call it something more provocative:
We outsourced coordination to the patient.
That's backwards.
The patient should not be the integration layer.
The Same Problem Exists in Revenue Cycle Management
The patient shouldn't be the integration layer.
Neither should the biller.
Yet billing staff often become the human middleware
connecting incompatible systems.
EHR.
Clearinghouse.
Payer portal.
Phone call.
Fax.
Email.
Spreadsheet.
Denial report.
Back to EHR.
The biller becomes the API.
Except the API takes lunch.
And gets tired.
And can quit.
And deserves better.
That isn't an insult to billing professionals.
It's an indictment of the systems we have built around them.
What AI Should Actually Do in Medical Billing
A useful AI billing system should be able to understand
workflow context.
Not just generate text.
It should help answer:
What happened?
Why did it happen?
What information matters?
What is missing?
What should happen next?
Who should handle it?
What requires human approval?
What happened after the action?
That is a workflow loop.
Not a chatbot.
Not a document generator.
A workflow loop.
The Future of Medical Billing Is Not “AI Billing”
I think the phrase itself is too small.
The future is workflow intelligence.
Billing is one example.
Prior authorization is another.
Referral management.
Documentation.
Credentialing.
Patient communication.
Care coordination.
Quality reporting.
Each contains repetitive cognitive work.
The opportunity isn't to replace the human being.
It's to remove the unnecessary work surrounding the human
being.
Where We Should Be Careful
There are legitimate risks.
AI can hallucinate.
AI can misclassify.
AI can amplify bias.
AI can expose protected health information if deployed
irresponsibly.
AI can make incorrect coding recommendations.
AI can create false confidence.
AI can automate bad decisions at scale.
And AI can create a particularly dangerous illusion:
“The computer said so.”
That is not accountability.
Healthcare organizations need appropriate privacy and
security safeguards, contractual protections where applicable, auditability,
human oversight, validation, monitoring and clear responsibility for decisions.
The more consequential the action, the stronger the controls
should be.
Automation should not become a way to make responsibility
disappear.
The Ethical Question Nobody Likes
Here's a question every healthcare AI company should answer:
Who benefits from the automation?
Does the physician benefit?
Does staff benefit?
Does the patient benefit?
Does the practice benefit?
Does the payer benefit?
Or does the technology simply allow one organization to push
more work onto another organization faster?
That's not innovation.
That's burden transfer.
And healthcare already has enough of that.
A Warning About “Efficiency”
Efficiency is not automatically good.
Suppose AI allows a billing department to process twice as
many claims.
Excellent.
But suppose it also doubles the number of claims entering an
already dysfunctional payer workflow.
Not so excellent.
Suppose an AI tool increases documentation completeness.
Great.
But suppose payers respond with aggressive downcoding.
Now the system has entered an arms race.
This is exactly why recent analysis of healthcare AI has
warned that applying automation to fragmented administrative systems can
produce an efficiency paradox rather than true efficiency.
The lesson:
Optimize the system, not just the task.
Recent News Is Pointing in the Same Direction
The healthcare AI conversation is shifting.
It is no longer simply:
“Can AI work?”
The more important questions are becoming:
Does it work inside real workflows?
Does it save time?
Does it reduce burden?
Can physicians trust it?
Does it create new administrative work?
Who remains accountable?
The AMA's 2026 research shows AI use among physicians has
reached 81%, while its research also emphasizes responsible implementation and
the need for physician involvement.
At the same time, healthcare reporting is increasingly
questioning whether AI deployed on top of fragmented administrative processes
can actually worsen complexity.
That tension will define the next phase of healthcare AI.
The Biggest AI Opportunity May Be Boring
This may disappoint the people building futuristic
healthcare demos.
But I think one of the biggest opportunities in healthcare
AI is incredibly boring.
Find the missing document.
Find the reason for the denial.
Find the next step.
Find the person responsible.
Find the information already sitting somewhere in the
system.
Remove the duplicate task.
Reduce the handoff.
Prevent the error.
Close the loop.
That's not science fiction.
It's operational intelligence.
And healthcare desperately needs it.
What Lois Bockmann's Story Ultimately Teaches Us
Bockmann's story is not a story about technology replacing
clinicians.
It is almost the opposite.
It is a story about what becomes possible when people have
enough attention and coordination to see the whole picture.
Bockmann needed someone to listen to the symptom that didn't
fit neatly into the existing narrative.
She needed clinicians who could connect information.
She needed specialists who communicated.
She needed a team that could coordinate two very different
surgical problems.
Technology may support that process.
But technology wasn't the protagonist.
People were.
That distinction matters.
Because the purpose of healthcare technology shouldn't be to
make healthcare feel more technological.
It should make healthcare feel more human.
My Contrarian Prediction
Over the next decade, healthcare will not be transformed by
the organization with the most AI.
It will be transformed by the organizations that understand where
AI belongs — and where it doesn't.
The winners won't necessarily have the biggest models.
They'll have the best workflows.
They'll know which decisions require physicians.
Which tasks require nurses.
Which work can be automated.
Which information needs to move.
Which exceptions need escalation.
And which processes should simply be eliminated.
The smartest healthcare organization may therefore be the
one that does the least unnecessary work.
That's a very different definition of innovation.
The Question I Would Ask Every Practice Owner
Forget the AI roadmap for a moment.
Forget the vendor demonstrations.
Forget the impressive PowerPoint.
Walk into your practice tomorrow and ask:
“What is everybody doing here that nobody should have to
do?”
Don't ask what can be automated.
Start there.
Find the waste.
Find the repetition.
Find the friction.
Find the unnecessary handoff.
Find the task that exists because two systems don't
communicate.
Find the task that exists because a payer designed it that
way.
Find the task that exists because “that's how we've always
done it.”
Then ask:
Can we eliminate it?
If not:
Can we simplify it?
If not:
Can AI assist with it?
And if AI assists:
Where should the human remain in control?
That sequence is more important than buying another AI tool.
Failure Is Part of the Process
Healthcare organizations should become more comfortable
admitting when technology doesn't work.
An AI tool may promise to save time and instead create
review work.
A chatbot may reduce phone calls but increase confusion.
An automated denial system may generate more appeals without
improving resolution.
A documentation tool may create better notes but introduce
coding disputes.
That's not necessarily evidence that AI is useless.
It may mean the workflow was poorly designed.
The answer isn't always:
“The AI needs improvement.”
Sometimes the answer is:
“We shouldn't be doing this task this way in the first
place.”
That's a harder answer.
It is also often the more valuable one.
Myth Buster
Myth 1: AI will eliminate medical billing.
Probably not.
It will change which parts require human effort.
Myth 2: More automation always means lower costs.
No.
Poorly designed automation can create new work.
Myth 3: AI is primarily a clinical tool.
Not necessarily.
Physicians themselves identify administrative work as one of
AI's biggest opportunities.
Myth 4: Human review means the AI isn't intelligent.
Wrong.
Human oversight can be a feature of responsible automation.
Myth 5: Burnout is mainly a physician resilience problem.
No.
Workflow design, documentation burden, operational
reliability and administrative friction matter.
Myth 6: Small practices can't benefit from AI.
They can — provided the technology reduces complexity rather
than adding another layer.
Practical Checklist for Practice Leaders
Before implementing any AI workflow, ask:
1. What problem are we actually solving?
If the answer is “we want AI,” stop.
2. What does the current workflow look like?
Map it.
3. How many handoffs exist?
Count them.
4. Which steps require judgment?
Protect them.
5. Which steps are repetitive?
Candidate for automation.
6. What information does the system need?
Make sure it can access reliable data.
7. What happens when the AI is wrong?
Define escalation.
8. Who remains accountable?
Name the role.
9. How will we measure success?
Choose metrics before launch.
10. What work disappears?
This may be the most important question of all.
The Future Is Not Autonomous Healthcare
At least, that's not the future I want.
I don't want a healthcare system where humans disappear
behind software.
I want a healthcare system where humans spend more time
doing the things humans are uniquely good at.
Listening.
Reasoning.
Explaining.
Comforting.
Deciding.
Connecting.
Advocating.
Caring.
And where machines handle more of the work machines are good
at.
Retrieving.
Comparing.
Sorting.
Monitoring.
Routing.
Summarizing.
Detecting patterns.
Preparing work.
Closing loops.
That is not humans versus machines.
It is humans doing more human work because machines
handle more machine work.
The Real Innovation Is Giving Attention Back
That brings us back to Lois Bockmann.
She wasn't saved because someone built a prettier dashboard.
She wasn't saved because someone added another notification.
She wasn't saved because an algorithm generated a 500-word
summary.
Her story turned when people connected the pieces.
When someone listened.
When someone investigated.
When one specialist contacted another.
When a team coordinated.
When information became action.
That is the lesson I keep coming back to.
Healthcare doesn't just need better information.
It needs better attention.
And attention requires workflow.
About OnnX
I founded OnnX around a simple belief:
Healthcare technology should remove administrative
friction, not create another layer of it.
Our focus is on AI-powered workflow automation for medical
billing and revenue-cycle operations, particularly for small and medium-sized
medical practices.
The goal isn't to tell physicians that AI will replace their
staff.
It is to ask a more practical question:
Which repetitive cognitive tasks can technology safely
take off their plates?
That includes analyzing claims and denials, identifying
likely causes, organizing relevant information, recommending next steps,
preparing work for review and helping route the right task to the right person.
The larger idea is workflow intelligence.
Because the future of healthcare AI shouldn't simply be
about generating more information.
It should be about turning information into the right action
with less friction.
Tools and Metrics Worth Watching
For organizations beginning this journey, don't start with a
shopping list of AI vendors.
Start with measurement.
Track:
- Administrative
hours per physician
- Manual
touches per claim
- Denial
rate
- Clean
claim rate
- Days
in A/R
- Denial
resolution time
- Appeal
success rate
- Rework
rate
- First-pass
resolution
- Staff
time spent on repetitive tasks
- Number
of workflow handoffs
- Number
of unnecessary escalations
- Physician
administrative time
- Patient
wait time where applicable
Then establish a baseline.
If the number doesn't improve, the project isn't finished.
And if the technology doesn't improve it?
Be willing to say so.
Legal and Compliance Considerations
Healthcare automation has consequences.
Organizations should evaluate:
Privacy: How is protected health information handled?
Security: Who can access the system?
Contracts: Are appropriate agreements and business
associate arrangements in place where required?
Auditability: Can the organization determine what the
system did and why?
Human oversight: Which decisions require human
approval?
Coding: Does the recommendation align with
documentation and applicable coding requirements?
Payer rules: Does automation account for
payer-specific requirements?
Accountability: Who is responsible when the system
makes a mistake?
AI should never become a convenient place to hide
responsibility.
The Bigger Question
Maybe the healthcare industry's most important AI question
isn't:
“How intelligent can our systems become?”
Maybe it's:
“How much unnecessary work can we remove?”
That's a much harder question.
Because answering it may require admitting that some
workflows shouldn't exist.
Some approvals aren't necessary.
Some handoffs are redundant.
Some dashboards are noise.
Some reports are never used.
Some tasks exist because two organizations haven't agreed on
a better way.
Some work exists because everyone got used to doing it.
And some work exists because nobody has stopped to ask why.
AI can help us solve those problems.
But first we have to be willing to see them.
Final Thought
Lois Bockmann spent her career helping other people navigate
healthcare.
Then she became the person navigating it.
Her story is a reminder that being inside the system
doesn't guarantee that the system will see you clearly.
It also reminds us that coordination isn't an administrative
luxury.
It can be clinical.
Attention isn't a soft skill.
It can be consequential.
Workflow isn't merely an operations problem.
It can shape outcomes.
And technology isn't valuable simply because it is
intelligent.
It is valuable when it helps people do the right thing at
the right time with less unnecessary friction.
Maybe that is the real opportunity for AI in healthcare.
Not more screens.
Not more dashboards.
Not more notifications.
Not more tools competing for attention.
Less noise.
Less repetition.
Less administrative friction.
More attention where it matters.
Because the patient is not the workflow.
The patient is the reason the workflow exists.
And perhaps the ultimate measure of healthcare innovation is
surprisingly simple:
Did we give the human being more room to care for another
human being?
Continue the Conversation
I write and speak about the intersection of medicine, AI,
healthcare operations, workflow automation, medical billing and the future of
healthcare.
You can continue the conversation through:
A free resource is also available through the Featured
section of Dr. Cham's LinkedIn profile. No signup is required.
References
1. WakeMed — “Coordinated Care at WakeMed Saved Nurse
Lois Bockmann's Heart — and Her Life.”
The real-life patient story behind this article, including Bockmann's
diagnoses, treatment and multidisciplinary care team. Read
the WakeMed story
2. American Medical Association — “AMA: AI usage among
doctors doubles as confidence in technology grows.”
The AMA's 2026 research showing that 81% of physicians report using AI
professionally and examining how physicians view its benefits and risks. Read
the AMA research summary
3. Medical Economics — “Why AI may be making your
administrative burden worse.”
An important counterpoint examining how AI layered onto fragmented
administrative processes can potentially increase complexity rather than
eliminate it. Read
the Medical Economics analysis
Disclaimer
This article is provided for general educational and
informational purposes only. It is not intended to constitute medical, legal,
regulatory, compliance, financial or professional advice. Healthcare
organizations, clinicians and businesses should evaluate their individual
circumstances and obtain appropriate advice from qualified professionals before
making clinical, operational, technological or financial decisions. References
to healthcare professionals, organizations, technologies or real-world cases are
provided for educational context and should not be interpreted as endorsements
unless explicitly stated.
Your Turn
If you could permanently eliminate one administrative
task from your medical practice tomorrow, what would it be?
The task that makes you think:
“Why are humans still doing this?”
I'd genuinely like to know.
Comment below.
If this story made you think differently about AI,
workflow or administrative burden, share it with a physician, practice
administrator, healthcare leader or colleague who is dealing with the same
problem.
Because perhaps the next great healthcare innovation isn't
another technology.
Perhaps it's finally removing the work that never needed
to exist.
Knowledge drives progress. Better questions drive
innovation. And sometimes the most important question in healthcare is simply:
Why are we still doing this?
#HealthcareAI #MedicalBilling #MedicalPractice #Physicians
#ClinicOwners #HealthcareInnovation #RevenueCycleManagement #RCM
#HealthcareAutomation #MedicalTechnology #AdministrativeBurden
#PhysicianBurnout #HealthTech #AIinHealthcare #IndependentPractice #PracticeManagement
#WorkflowAutomation #PatientCare #HealthcareLeadership #DigitalHealth
#HealthcareEntrepreneur #MedicalPracticeManagement #PhysicianEntrepreneur
#HealthcareOperations
