What One Rochester Nurse’s Story Reveals About the Human Cost of a Healthcare System That Keeps Asking Humans to Fix Its Machines
“The two dimensions of human relationships and alignment
on patient values and agency deserve particular protection as AI advances.”
— Eric
Horvitz, MD, PhD, Chief Scientific Officer, Microsoft, September 24,
2026
On September 16, 2026, Clint Peterson, a 26-year-old
registered nurse, was at St. Marys Hospital in Rochester, Minnesota.
He was doing what nurses do.
Taking care of someone else.
Then everything changed.
According to a criminal complaint cited by Rochester-area
reporting, Jascon Jacoby Hollins, a 43-year-old man in the custody of
the Olmsted County Sheriff’s Office, was receiving medical care at St. Marys
when he allegedly broke free from a restraint and ran into a hospital hallway.
A detention deputy pursued him.
During the encounter, the deputy and Hollins collided with
Peterson.
Peterson fell and struck his head.
He suffered a skull fracture, traumatic brain injury and
brain bleed requiring surgery, according to reporting about his condition.
The nurse became the patient.
And suddenly, the person whose job was helping others needed
help himself.
That is the part of this story that stays with you.
Peterson had reportedly recently celebrated his one-year
review at Mayo Clinic. He was considering additional leadership
responsibilities, including work related to patient safety and patient
experience.
Then, in an instant, the future became uncertain.
His longtime friend Matt Wurst, who says he has known
Peterson since kindergarten, described the reversal in painfully simple terms:
Clint is a nurse.
Taking care of people is a huge part of who he is.
Now he is the one who needs care.
Peterson reportedly has needed assistance with mobility,
medication management and daily activities. His recovery remains uncertain. No
one knows exactly when he will live independently again or whether he will
return to nursing.
The criminal case surrounding the incident is separate from
Peterson's recovery. Hollins faces criminal charges, and the allegations in the
complaint should not be treated as adjudicated facts.
But Peterson's story raises a much larger question.
Not about crime.
Not about hospitals.
Not even about nursing.
About human attention.
Because healthcare has a strange habit.
We take some of the most highly trained people in society...
and then ask them to spend enormous amounts of time fixing
problems created by the systems surrounding them.
Maybe we should stop calling that productivity.
Healthcare Has a Productivity Problem. But Maybe Not the
One We Think.
For decades, healthcare has asked:
How do we make clinicians more efficient?
Better workflows.
Better templates.
Better dashboards.
Better software.
Better automation.
Better artificial intelligence.
Better revenue-cycle management.
Better everything.
And yet the work keeps expanding.
The inbox grows.
The task list grows.
The prior-authorizations grow.
The denials grow.
The documentation grows.
The portals multiply like rabbits.
And somewhere in the middle of all of it sits the physician.
Still seeing patients.
Still answering messages.
Still signing charts.
Still explaining things.
Still fixing things.
Still wondering why a system designed to help healthcare
seems to require so much help itself.
Here's the contrarian question:
What if the goal shouldn't be making healthcare workers
faster?
What if the goal should be making less work necessary?
That's a very different proposition.
The Healthcare Industry's Favorite Trick
We have become remarkably good at automating the
consequences of bad processes.
A claim is denied.
Build an AI denial-management tool.
A payer asks for documentation.
Build an AI authorization assistant.
A physician spends too much time documenting.
Build an AI scribe.
A staff member has to check three systems.
Build a dashboard.
A billing team has too many exceptions.
Build an exception-management platform.
A workflow is inefficient.
Add automation.
A workflow is still inefficient.
Add more automation.
Eventually we have a beautifully automated mess.
That is not a joke.
It is one of the central risks of healthcare AI.
AI is extraordinarily good at doing what we ask.
If we ask it to process unnecessary work faster, it can do
that too.
Artificial intelligence does not automatically produce
intelligent systems.
Sometimes it simply produces faster bureaucracy.
The Rube Goldberg Revenue Cycle
Consider a typical revenue-cycle problem.
A patient arrives.
Information is collected.
Someone enters it.
Another system receives it.
Someone interprets it.
Someone checks eligibility.
Someone checks authorization.
Someone documents the encounter.
Someone codes it.
Someone submits the claim.
The payer processes it.
Something doesn't match.
The claim is rejected.
Someone investigates.
Someone opens a portal.
Someone calls the payer.
Someone sends documentation.
Someone asks the physician a question.
The physician stops what they're doing.
Someone resubmits the claim.
Someone waits.
Someone checks again.
Someone follows up.
Someone reconciles the payment.
Then everyone celebrates.
Why?
Because the claim finally got paid.
Congratulations.
We successfully completed a process that required twelve
people to repair information that should have been correct much earlier.
And then we call that revenue-cycle management.
Sometimes it feels more like revenue-cycle archaeology.
The Problem May Not Be Billing
This is where I disagree with a lot of conventional RCM
thinking.
The denial is not necessarily the problem.
The denial may be the symptom.
The billing department may not be the origin.
The claim may not be the origin.
The payer may not even be the origin.
The problem may have started much earlier.
At the point where information was captured.
At the point where a workflow allowed ambiguity.
At the point where two systems interpreted the same
information differently.
At the point where a person had to remember something a
system should have known.
At the point where someone entered the same information
twice.
The farther downstream the problem travels, the more
expensive it becomes to fix.
A five-second correction upstream can become a 30-minute
administrative event downstream.
And a 30-minute event repeated 500 times isn't a small
problem.
It's a business model.
The Human Attention Tax
We tend to calculate healthcare costs in dollars.
We should.
But dollars are not the only scarce resource.
Attention is scarce too.
A physician has a finite amount.
A nurse has a finite amount.
A medical assistant has a finite amount.
A practice manager has a finite amount.
A billing specialist has a finite amount.
And unlike software licenses, human attention does not renew
at midnight.
Every interruption consumes some of it.
Every context switch consumes some.
Every ambiguous request consumes some.
Every exception consumes some.
Every missing piece of information consumes some.
And sometimes the cost is invisible.
A physician spends 15 minutes fixing an administrative
issue.
The spreadsheet says:
15 minutes.
But the real cost may include:
15 minutes of direct time.
plus interruption.
plus cognitive switching.
plus delayed work.
plus another task pushed later.
plus the possibility of taking unfinished work home.
That is why I think healthcare needs a new metric:
Return on Attention.
Not just return on investment.
Not just productivity.
Return on Attention.
Where did the human attention go?
Was it necessary?
Could it have been prevented?
Could software have handled it?
Could better information have prevented it?
Could the human have spent that time with a patient instead?
Those questions may tell us more about healthcare technology
than another dashboard showing utilization.
Eric Horvitz Just Made the Human Point
On September 24, Microsoft Chief Scientific Officer Eric
Horvitz, MD, PhD, spoke about AI and healthcare at Washington University in
St. Louis.
His warning was not that AI should replace people.
Quite the opposite.
He emphasized protecting human relationships, patient
values and patient agency as AI advances. He described a future in which AI
assists clinician teams while human needs remain central.
That is important.
Because the conversation around AI can become strangely
mechanical.
We ask:
How many tasks can AI perform?
How many employees can AI replace?
How many minutes can AI save?
How many claims can AI process?
Those are useful questions.
But they aren't the whole question.
A better one is:
What should humans do with the attention AI gives back?
If the answer is “more administrative work,” we missed the
point.
AI Should Not Become the World's Fastest Intern
Here's another uncomfortable thought.
Healthcare may be at risk of building AI systems that
function like extremely energetic interns.
They never sleep.
They never complain.
They process enormous amounts of information.
And then they create 47 tasks for someone to review.
Fantastic.
Now the physician has an AI-generated inbox.
Progress?
Maybe.
But only if the human workload actually falls.
Otherwise we've created a new category of work:
checking the machine that checked the machine.
At some point, someone has to ask:
Who is actually getting the time back?
The Best Automation May Be the Automation Nobody Notices
The most impressive AI demo isn't necessarily the most
valuable AI system.
A system that generates a spectacular summary is impressive.
A system that eliminates the need for the summary may be
more valuable.
A system that identifies a denial quickly is useful.
A system that prevents the denial is potentially more
valuable.
A system that routes an exception efficiently is useful.
A system that makes the exception unnecessary is a different
class of improvement.
This is the distinction between:
automation of work
and
elimination of work.
We should stop treating them as the same thing.
Healthcare's Most Expensive Phrase
There is a phrase every practice owner should question:
“That's just how we do it.”
Those six words have probably cost American healthcare
billions.
Because work that starts as a workaround can eventually
become:
a workflow,
then a policy,
then a habit,
then a software requirement,
then a job description,
then a department,
then an entire industry.
Nobody remembers why the process exists.
Everyone just knows someone has to do it.
This is how organizational folklore becomes infrastructure.
The Physician as Human Middleware
Here's one of the strangest roles physicians have acquired.
Human middleware.
Information doesn't move cleanly between systems.
So the physician becomes the translator.
The payer wants something.
The physician explains.
The coder needs clarification.
The physician explains.
The authorization department needs documentation.
The physician explains.
A claim has a problem.
The physician explains.
The EHR doesn't understand the context.
The physician supplies the context.
The physician becomes the API.
Except the API has a medical degree, a patient schedule and
approximately 37 unread messages.
That is not a scalable architecture.
The Real Cost of a Physician Touch
Let's make this concrete.
Suppose a claim requires physician intervention.
Maybe it takes three minutes.
Sounds harmless.
Now multiply it.
One physician.
Twenty claims.
Five days.
Four weeks.
One year.
The number becomes meaningful.
Now imagine the physician is interrupted rather than
handling the work in a dedicated block.
The cognitive cost increases.
Now imagine the issue could have been prevented upstream.
The question changes again.
We shouldn't only measure:
How fast did the physician resolve it?
We should measure:
Why did the physician have to touch it?
That is a much more powerful operational question.
Five Metrics I'd Like to See More Practices Track
1. Physician Touches Per Claim
How many claims require physician intervention for
administrative reasons?
Not clinical judgment.
Administrative reasons.
Track it.
2. Manual Touches Per Claim
How many people have to touch the transaction?
Every handoff is an opportunity for information loss.
3. Rework Rate
How much work is being repeated because something wasn't
correct the first time?
4. Exception Rate
How often does the standard process fail?
5. Administrative Minutes Per Encounter
How much human time surrounds a patient encounter that isn't
direct patient care?
These metrics won't solve the problem.
But they expose it.
And you can't redesign what you can't see.
A Seven-Day Experiment for Your Practice
You don't need another software platform.
Start with a spreadsheet.
For seven days, capture every recurring administrative
interruption.
Write down:
What happened?
Who discovered it?
When was it discovered?
Who touched it?
Did the physician become involved?
What information was missing?
Where was that information originally available?
Could the problem have been prevented earlier?
What was the downstream consequence?
Don't fix anything yet.
Just observe.
Because healthcare has a bad habit of solving the first
visible problem.
The better approach is to find the earliest preventable
cause.
The Upstream Test
Here's a simple rule:
When you find a problem, move backward.
A claim was denied.
Move backward.
Why?
Missing information.
Move backward.
Why was it missing?
It wasn't captured.
Move backward.
Why wasn't it captured?
The workflow didn't require it.
Move backward.
Why?
The system assumed someone would know.
Move backward.
Why?
Because that's how the practice has always done it.
There it is.
The real problem may not have been the denial.
It may have been a design assumption made months or years
earlier.
This Is Where OnnX Starts
This is the thinking behind OnnX.
The idea is simple:
Most of the problem starts upstream.
Healthcare billing is often treated as a downstream workflow
problem.
I think a significant part of it is an information-structure
problem.
If clinical and operational information is inconsistent,
incomplete or fragmented at the beginning, downstream systems have to
compensate.
Someone has to interpret it.
Someone has to correct it.
Someone has to chase it.
Someone has to resubmit it.
Someone has to explain it.
The system becomes increasingly reactive.
OnnX is built around a different premise:
Structure the information earlier.
Reduce variability before it becomes a billing problem.
That doesn't mean every denial disappears.
It doesn't mean every payer rule becomes simple.
And it certainly doesn't mean AI magically understands
healthcare.
It means we should attack preventable variability closer to
its source.
We're Optimizing Around Noise
Healthcare has become incredibly sophisticated at responding
to noise.
We classify it.
Route it.
Prioritize it.
Analyze it.
Escalate it.
Automate it.
Report it.
Measure it.
And then we congratulate ourselves for managing the noise.
But what if the noise is telling us something?
Maybe the system isn't signaling that we need a better
noise-management department.
Maybe it's signaling that the information architecture is
wrong.
We are often optimizing around noise instead of removing
it.
That distinction is central to the future of healthcare AI.
AI Is Only as Good as the Problem You Give It
There is a seductive idea that AI will simply make
healthcare more efficient.
It might.
But AI doesn't determine the problem.
Humans do.
If we give AI a bad process, it may optimize the bad
process.
If we give it fragmented information, it may organize
fragmented information.
If we give it ambiguous instructions, it may produce very
sophisticated ambiguity.
That means the most important AI skill may not be prompting.
It may be problem definition.
Before asking:
“What can AI do?”
Ask:
“What should no human have to do anymore?”
That's the better starting point.
Three Experts, Three Lessons
Eric Horvitz, MD, PhD: Protect the Human Relationship
Horvitz's September 24 remarks emphasize human
relationships, patient values and agency as AI becomes more capable.
The lesson:
Don't confuse intelligence with understanding.
A system can process information without understanding what
matters to a patient.
Healthcare still needs people who listen.
People who interpret.
People who explain.
People who take responsibility.
AI should strengthen those functions, not bury them.
Willie Underwood III, MD, MSc, MPH: Patients Shouldn't
Have to Fight the System
Earlier this month, AMA President Willie Underwood III,
MD, MSc, MPH, wrote that patients should not have to fight the healthcare
system to receive care they need. His comments focused on Medicare payment and
prior authorization, but the broader principle is relevant: administrative
friction eventually reaches the patient.
The lesson:
Administrative friction isn't merely an employee problem.
It can become a patient problem.
When information doesn't move, care can slow.
When authorization doesn't move, treatment can slow.
When claims don't move, patients and practices can spend
time chasing the consequences.
The back office is not disconnected from the exam room.
It is connected by every piece of information moving between
them.
Bruce Swords, MD: AI Can Give Physicians Time Back — If
We Use It Correctly
In an interview published September 24, Bruce Swords, MD,
chief clinical officer at Bon Secours St. Francis, described AI as useful for
administrative work historically handled by physicians.
But he also emphasized the continuing importance of direct
clinical oversight and human context.
The lesson:
AI should reduce administrative burden without pretending
to replace clinical judgment.
That distinction is critical.
Good automation doesn't make physicians less important.
It makes unnecessary physician work less important.
Recent News Is Making the Same Point From Different
Directions
Something interesting is happening across healthcare right
now.
A family in Massachusetts recently described repeatedly
resubmitting a medical insurance claim after it was denied over missing
information. The Connecticut Office of the Health Care Advocate said 65% of
calls to its office involve denials, including administrative denials that
essentially require people to “connect the dots.” The insurer ultimately
acknowledged confusion and approved the claim.
That story sounds small.
It isn't.
Because healthcare is full of tiny administrative failures.
One missing code.
One missing field.
One missing document.
One misunderstood requirement.
One incorrect assumption.
Multiply those events across millions of encounters.
Suddenly the “small” problem becomes a system.
And then we hire people to manage the system.
The Billion-Dollar Question
Here's the question I would ask every healthcare technology
company:
How much work did your product eliminate?
Not automate.
Not accelerate.
Not route.
Not dashboard.
Eliminate.
If your software saves a billing employee 20 minutes but
creates 15 minutes of review, the net gain is five minutes.
If AI creates a draft that requires another person to
inspect every line, that's not zero work.
If an automated system generates ten alerts and someone has
to review all ten, you've moved the work.
Maybe that's still valuable.
But let's call it what it is.
Work relocation is not work elimination.
What Physicians Should Ask Vendors
The next time someone shows you an impressive AI demo, ask:
What happens when the AI is wrong?
Then ask:
Who reviews it?
Then:
How many new tasks does that create?
Then:
What information does the system need?
Then:
Where does that information come from?
Then:
What happens upstream if that information isn't
available?
And finally:
How much human work disappears?
That last question should be on every healthcare technology
RFP.
What Clinic Owners Should Ask Their Own Teams
Ask:
What do you do every day that you think is ridiculous?
Not “inefficient.”
Not “suboptimal.”
Ridiculous.
People know.
They may not say it in meetings.
They know which spreadsheet nobody trusts.
They know which payer portal they hate.
They know which report gets exported and immediately
re-entered somewhere else.
They know which fax gets sent because “that's what they
require.”
They know which task exists because somebody five years ago
said, “Just in case.”
That institutional knowledge is valuable.
Use it.
The Ethical Question: What Happens to the Work We Remove?
There is a legitimate concern here.
If AI eliminates repetitive administrative work, what
happens to the people who perform it?
That question deserves a serious answer.
The goal should not simply be:
Replace people.
A better goal is:
Remove unnecessary work and redirect human capacity.
People can handle:
- complex
exceptions
- patient
communication
- problem-solving
- relationship
management
- quality
improvement
- clinical
support
- coordination
- judgment
The ethical question is not whether work changes.
It will.
The question is whether the transition produces more
human value or simply fewer humans.
Technology is not inherently humane.
Its design and deployment determine the outcome.
The Legal and Compliance Reality
Upstream automation also creates responsibility.
Healthcare organizations still need appropriate controls
around:
HIPAA and privacy
Security
Access controls
Business associate relationships
Documentation
Coding accuracy
Claims integrity
Audit trails
Human oversight
Data provenance
Model governance
Automation doesn't transfer responsibility to the algorithm.
“AI did it” is not a compliance strategy.
A trustworthy system should make it possible to answer:
What information was used?
Where did it come from?
What happened to it?
What rule or model influenced the action?
Who approved it?
What changed?
Can the organization reconstruct the decision?
The future of healthcare AI needs not only intelligence.
It needs traceability.
What We Should Stop Measuring
Maybe we need fewer vanity metrics.
Not fewer measurements.
Better ones.
Stop celebrating automation simply because it exists.
Stop celebrating task completion without asking whether the
task should exist.
Stop celebrating claims processed without measuring rework.
Stop celebrating denial resolution without measuring
avoidable denials.
Stop celebrating staff utilization without measuring
unnecessary touches.
Stop celebrating AI adoption without measuring human
attention returned.
The important metric is not:
How much technology did we deploy?
It is:
What became unnecessary because we deployed it?
The Future of RCM May Be Less RCM
That sounds strange.
But consider it.
If information is structured earlier...
If fewer errors occur...
If fewer claims require intervention...
If fewer denials occur...
If fewer people have to touch the same transaction...
If fewer physician interruptions happen...
Then the organization needs less downstream repair.
That's not necessarily a better billing department.
It's a smaller problem.
And smaller problems are usually easier to manage.
The best revenue cycle may eventually be the one that gives
people less revenue-cycle work to do.
The Human Story Matters More Than the Technology
This is why Clint Peterson matters.
Not because his injury proves anything about medical
billing.
It doesn't.
Not because his story should be exploited as a metaphor for
AI.
It shouldn't.
He is a real person recovering from a serious injury.
But his story reminds us of something easy to forget:
The people who work inside healthcare are human beings
before they are job titles.
A nurse can become a patient.
A physician can become a patient.
A practice manager can become a patient.
A biller can become a patient.
And one day, every person working inside the system will
stand on the other side of it.
That changes the question.
We aren't designing workflows for machines.
We are designing them for people who may eventually need the
care those workflows support.
Imagine Getting 20% of Your Attention Back
Imagine a practice where 20% of recurring administrative
interruptions disappeared.
Not moved.
Not outsourced.
Gone.
What would happen?
Maybe physicians would finish clinic earlier.
Maybe nurses would have more time with patients.
Maybe staff would spend less time on repetitive work.
Maybe managers would have time to improve the practice
instead of constantly rescuing it.
Maybe patients would get faster answers.
Maybe nobody would notice at first.
And that's okay.
The best infrastructure often becomes invisible.
Nobody compliments the plumbing when the water comes out.
Nobody celebrates a bridge because traffic crossed it
without incident.
Nobody says:
“Wow, what an amazing absence of administrative friction.”
But that's the point.
Good infrastructure disappears into the background.
Three Things to Do Monday Morning
1. Find the work everyone has normalized.
Ask your team:
“What task do we do every week that makes no sense?”
Listen.
Don't defend the process.
2. Trace the problem upstream.
Don't stop at the denial.
Don't stop at the rejected claim.
Don't stop at the phone call.
Find the earliest point where better information could have
prevented the problem.
3. Measure attention returned.
Count:
physician minutes
staff minutes
manual touches
handoffs
rework
exceptions
Then ask:
How much of that disappeared?
That's your real automation ROI.
A Different Definition of Innovation
Healthcare doesn't need another technology arms race.
It needs better questions.
Instead of:
How can we automate this?
Ask:
Why does this exist?
Instead of:
How can we process it faster?
Ask:
Why are we processing it at all?
Instead of:
How can AI fix this?
Ask:
What information would prevent it?
Instead of:
How many tasks did we automate?
Ask:
How many tasks disappeared?
Instead of:
How productive is our staff?
Ask:
How much unnecessary work are we asking them to absorb?
That's a different philosophy of healthcare technology.
And I think it is overdue.
Don't Automate the Mess
Clint Peterson's story began with a sudden reversal.
The person who cared for others became the person who needed
care.
It is a reminder that human attention is not an unlimited
resource.
We should spend it carefully.
Healthcare workers shouldn't have to spend their best
cognitive energy repairing systems that could have been designed better.
Physicians shouldn't be the fallback database.
Nurses shouldn't have to become information detectives.
Practice managers shouldn't have to become professional
firefighters.
Billers shouldn't have to repeatedly reconstruct information
that should have been structured earlier.
And AI shouldn't become another layer of complexity sitting
on top of complexity.
The opportunity is bigger.
Don't automate the mess.
Don't optimize the noise.
Don't make humans faster at unnecessary work.
Remove the reason the work exists.
Then give the attention back to healthcare.
Frequently Asked Questions
Is all administrative work bad?
No.
Healthcare requires documentation, compliance, coordination
and financial operations.
The issue is avoidable administrative work.
Does upstream optimization mean eliminating billing
departments?
No.
Billing professionals still handle complex cases, payer
interactions, exceptions and judgment-intensive work.
The goal is to reduce preventable rework.
Is AI enough?
No.
AI is a tool.
Better information architecture, process design, governance
and accountability are equally important.
Should physicians trust AI?
Physicians should evaluate AI based on its intended use,
evidence, reliability, integration, oversight and risk.
AI should support appropriate clinical judgment rather than
replace it.
What should a small practice measure first?
Start with:
physician administrative minutes
manual touches
rework
exceptions
avoidable denials
Those numbers can reveal where the system is consuming human
attention.
What is OnnX trying to change?
OnnX focuses on the idea that many downstream billing
problems originate upstream in how clinical and operational information is
captured and structured.
The objective is not simply to manage more billing work.
It is to reduce the amount of preventable work that reaches
the revenue cycle in the first place.
Myth Busters
Myth: More automation always means less work.
Reality: Poorly designed automation can create new review
and exception work.
Myth: A denial is simply a billing problem.
Reality: The cause may originate much earlier in the
information chain.
Myth: More staff automatically fixes inefficiency.
Reality: Additional staff can add capacity while leaving the
underlying process unchanged.
Myth: AI makes bad information good.
Reality: AI can organize, summarize or process bad
information without correcting its underlying quality.
Myth: The physician should resolve administrative
exceptions.
Reality: Some require clinical judgment. Many do not.
Myth: Technology is transformation.
Reality: Technology is an instrument. Transformation occurs
when the system itself improves.
Final Thought
Clint Peterson went to work as a nurse.
Then he became the patient.
His story reminds us that the people inside healthcare are
the same people who eventually depend on healthcare.
That should change how we think about technology.
The goal shouldn't be to make people endlessly more
efficient at absorbing system failure.
The goal should be to reduce the amount of system failure
they have to absorb.
That is a different vision.
A quieter one.
Less glamorous, perhaps.
But much more useful.
Because the future of healthcare may not belong to the
system that automates the most.
It may belong to the system that makes the most
unnecessary work disappear.
And when that happens, something important comes back.
Time.
Attention.
Judgment.
Human connection.
Maybe that is what healthcare technology should have been
giving us all along.
Get Involved
Here's my question for physicians and clinic owners:
What is one administrative task in your practice that
everyone accepts as “normal” even though nobody would design it that way today?
Tell me in the comments.
Not the symptom.
The actual task.
And if you know another physician or clinic owner who is
spending valuable time fixing work that shouldn't exist, share this with
them.
Because maybe the next breakthrough in healthcare won't be
another AI tool.
Maybe it will be the decision to stop building tools around
broken processes.
About the Author
Dr. Daniel Cham is a physician, entrepreneur and
healthcare technology founder focused on the intersection of clinical practice,
healthcare operations and medical billing.
LinkedIn: Connect with Dr.
Daniel Cham
Website: Dr.
Daniel Cham
Spotify: Listen to the
podcast
YouTube: Watch
on YouTube
X: Follow on X
Facebook: Follow on
Facebook
Through OnnX, he is exploring how AI and better information
architecture can reduce unnecessary administrative work for physician-owned and
independent medical practices.
His central premise is simple:
Healthcare should spend more human attention on patients
and less on repairing preventable system friction.
Disclaimer
This article is intended for general educational and
informational purposes. It is not legal, medical, financial or compliance
advice. Healthcare organizations should consult appropriately qualified
professionals regarding their specific circumstances.
The discussion of the September 16, 2026 incident
involving Clint Peterson reflects currently reported information. Allegations
described in a criminal complaint should not be treated as established facts
unless determined through the appropriate legal process.
Continue the Conversation
What healthcare problem have we become so accustomed to
that we no longer question whether it should exist?
That's the conversation worth having.
If this perspective resonates with you, share it with
another physician, clinic owner or healthcare operator.
And check the Featured section of my LinkedIn profile for
a free resource — no signup required.
References
- Galloway,
Aaron. “Rochester Nurse’s Life Changed in Seconds After St. Marys
Incident.” Rochester Now, September 23, 2026.
Primary source for the Clint Peterson story, including Peterson's injury, Matt Wurst's comments, St. Marys Hospital, Rochester, Minnesota, and the reported circumstances involving Jascon Jacoby Hollins.
Read the Rochester Now report - Shiloach,
Noa, and Tanvi Gorre. “Chief Scientific Officer of Microsoft Talks AI in
Healthcare.” Student Life, Washington University in St. Louis,
September 24, 2026.
Source for Eric Horvitz, MD, PhD's September 24 discussion of AI in healthcare and his warning that human relationships, patient values, and patient agency deserve particular protection as AI advances.
Read the Eric Horvitz interview/report - American
Medical Association. “AMA: Physician Burnout Rates Are Falling, Specialty
Gaps Remain.” April 16, 2026.
Source for the 2025 physician burnout figure of 41.9%, compared with 43.2% in 2024 and 48.2% in 2023. The data came from nearly 19,000 physician responses across 106 health systems and organizations.
Read the AMA burnout report - American
Medical Association. “Allocation of Physician Time in Ambulatory
Practice.”
Source for the finding that physicians spend nearly two hours on EHR and desk work for every hour of direct clinical face time during the clinic day, with another 1–2 hours of computer and clerical work outside office hours.
Read the AMA physician-time research - American
Medical Association. “Doctors Work Fewer Hours, but the EHR Still Follows
Them Home.”
Source for the 2024 physician-workweek data: 57.8 hours per week, including 27.2 hours of direct patient care, 13 hours of indirect patient care and 7.3 hours of administrative work.
Read the AMA physician-workweek report - American
Medical Association. “Delays in Reforming Medicare, Prior Authorization
Harm Patients.”
Source for Willie Underwood III, MD, MSc, MPH's discussion of administrative barriers and patient access to care. This is useful as a supporting reference, but I would not present Underwood's statement as the article's “this week” quote, since the article's current lead quote is now Eric Horvitz's September 24 statement.
Read the AMA article
#Healthcare #HealthcareAI #MedicalBilling #MedicalPractice
#PhysicianPractice #HealthTech #ArtificialIntelligence #RevenueCycleManagement
#PhysicianBurnout #HealthcareInnovation #AdministrativeBurden #DigitalHealth
#PhysicianEntrepreneur #AIinHealthcare #IndependentPractice
No comments:
Post a Comment