Insights for clinical leaders, physicians, nurses, and care teams focused on improving care delivery, reducing administrative burden, and enhancing the clinical experience.
Clinical documentation is changing quickly, and Ambient AI is a big reason why. It is helping providers reduce manual work, ease documentation burden, and move through visits more efficiently. But even as that shift accelerates, one reality remains true: a sizable percentage of providers still rely on dictation, and for many of them, that workflow is deeply ingrained in how they practice medicine.
Documentation should support how providers actually work
This is why I think the future of documentation has to account for how providers actually work today. Some clinicians are ready to fully embrace ambient workflows. Others want to dictate. Many will want both, depending on the visit, the setting, or simply personal preference. A modern documentation platform should support that full range.
The real opportunity is not choosing between dictation and ambient but bringing them together in a way that creates a more comprehensive experience. A provider might record a patient visit and let Ambient AI generate the first version of the note, then use dictation to add findings, make edits, or refine details for that encounter. In another case, a provider may skip recording altogether and dictate the note from start to finish. Both workflows should be supported in a comprehensive platform.
Having this flexibility matters because adoption does not happen in a vacuum. For some providers, moving from long-standing dictation habits to fully ambient documentation feels natural. For others, it feels like too much change all at once. Giving them access to dictation within the same broader platform lowers that barrier and lets organizations support more clinicians without forcing a one-size-fits-all transition.
AI can make dictation more valuable
This is also where AI starts to make documentation meaningfully better, even within dictation workflows. The value goes well beyond transcription. While a provider is dictating, AI can help choose the appropriate note template, automatically pull in relevant patient demographics, medical history, and lab data, reducing the amount of manual work needed to build a complete note. That saves time and helps the provider focus on the clinical thinking behind the encounter rather than the mechanics of documentation.
Over time, AI can help improve what is being documented, correct mistakes in real time, and connect providers to deeper intelligence while they work. For instance, a clinician could pause during dictation and ask an AI assistant a question about the patient’s history, get reminded of key details from a complex treatment course, or surface information that might otherwise require digging through the chart. In the right workflow, AI can also support clinical decision-making by helping providers quickly access evidence-based treatment recommendations in real time. The system could also help catch missing elements, identify documentation gaps, or prompt providers to include details that affect compliance, coding, and reimbursement.
When dictation sits inside a connected AI platform, it becomes part of a smarter documentation ecosystem that can work alongside ambient capture, AI assistants, and downstream workflows in a way that standalone tools cannot.
The goal is a more complete path forward
I do think the industry will continue moving toward broader AI adoption in documentation. Over time, more providers will shift toward ambient workflows because they open the door to much more advanced capabilities. But that future will come faster if we acknowledge where clinicians are today and give them the tools that fit into real practice patterns right now.
The goal should be simple: give every provider a documentation experience that works for their specialty, their setting, and their workflow. When you do that, you are building a more complete path forward for clinical documentation. This is exactly what we offer at Commure by rolling out dictation directly inside our Ambient AI solution.
Ambient scribing gets most of the attention when clinicians talk about AI in the exam room. And for good reason, it's one of the most tangible, immediately impactful technologies to ever enter clinical practice. But there's a quieter, equally significant problem that AI is now starting to solve: what happens before a provider ever walks into the room.
Pre-charting. And for most clinicians, it's a significant, largely invisible drain on their day.
What Is Pre-Charting, and Why Does It Matter?
Before a provider sees a patient, the provider prepares by reviewing previous visit notes, checking recent labs, scanning the problem list, and looking through active medications. The goal is to walk into the encounter fully informed, understanding what's changed, what needs follow-up, and what the priorities are for this patient.
Done well, pre-charting enables better care. A physician who's done their homework before stepping into the room can be more present with the patient, ask more targeted questions, and make more confident clinical decisions. The preparation is the foundation for everything that follows.
The problem is that it takes time. A lot of it.
In a typical outpatient setting, pre-charting for even simple patient encounters can take 10 minutes or more. For a panel of 20 or 30 patients in a day, those minutes can compound into hours. In more complex care settings (oncology, for instance), pre-charting a single patient can take up to 30 minutes! Patients may have long and complex histories, extensive labs, imaging across multiple time points, and notes from multiple specialists. Getting up to speed on where a patient stands before a visit requires synthesizing an enormous amount of information scattered across a chart.
And just like documentation at the end of the day, pre-charting often spills outside of clinical hours. Physicians are doing it at home the night before and reviewing charts in the early morning before their first patient. The same "pajama time" problem that ambient scribing addresses on the documentation side exists on the pre-charting side too, it's just less visible, and less discussed.
The Limits of the EHR for Pre-Charting
Part of what makes pre-charting so time-consuming is the nature of the tools clinicians are using to do it.
Electronic health records were designed primarily as documentation and data storage systems. They were not built to synthesize clinical information on demand. A provider trying to quickly understand a patient's current clinical status has to navigate across multiple sections (notes, labs, imaging, problem lists, medications) and do the synthesis themselves. The EHR surfaces the data. The clinician does all the interpretive work.
This is a structurally inefficient workflow. It puts the burden of information aggregation on the clinician, whose strength is clinical judgment, not data retrieval. It's the same fundamental problem that makes documentation itself so burdensome: technology that forces providers to operate at the bottom of their license rather than the top.
What a Pre-Visit Summary Actually Does
A pre-visit summary is automated pre-charting. Rather than requiring a provider to navigate through a chart and assemble the clinical picture manually, the technology pulls together the relevant information and surfaces it in a structured, readable format.
Commure’s pre-visit summary functionality is made possible by our Patient360 platform, the clinical data foundation that makes AI useful at the point of care. Patient360 unifies a patient's full EHR history into a structured, queryable data object, surfacing specialty-tailored information, source-cited AI Q&A, and longitudinal patient context at the moment it matters most.
At its best, a pre-visit summary gives a provider (before they ever enter the room) a coherent view of:
Patient at a glance: why this patient is being seen, what makes them clinically distinctive, what the key active issues are
Last visit summary: what happened at the previous encounter, what was assessed and planned, what follow-up was ordered
Today's focus: relevant recent developments, outstanding labs or results, issues flagged for this visit
Active diagnoses and medications: current problem list and medication list in a reviewable format, including when medications were started
Social history and relevant context: information that shapes clinical decision-making but often takes time to locate in a standard chart view
This is the information a clinician is hunting for during pre-charting. A pre-visit summary delivers it in one place, automatically, before the visit begins.
The Hallucination Problem and Why It Has to Be Solved
Any clinician hearing about AI-generated clinical summaries has an immediate and legitimate concern: what if the system hallucinates a part of the history?
In clinical medicine, a hallucinated finding or fabricated lab result is a patient safety and medicolegal issue. A provider who sees inaccurate information in a pre-visit summary and relies on it in clinical decision-making is in a worse position than a provider who had to do pre-charting manually.
This is why any serious pre-visit summary technology has to solve the attribution problem, not just the synthesis problem. It's not enough to generate a summary that reads well. Every clinical claim in that summary needs to be traceable to a specific source in the patient's chart so the provider can verify the information, understand its provenance, and calibrate their confidence accordingly.
Expandable citations that link directly to the source documentation (the specific note, the specific lab result, and the specific imaging report) are what make pre-visit summaries trustworthy. The AI does the synthesis. The clinician retains the ability to verify.
Turning Pre-Visit Summaries Into Clinical Intelligence
For patients with complex, longitudinal histories (chronic illness, oncology, or multi-system disease to name a few), even a well-constructed summary may leave questions unanswered. A provider might want to know what happened to a specific lab value over the past year, or which specialist last addressed a particular issue, or whether a medication change in the last visit was associated with any documented side effects.
When a pre-visit summary is built on top of a comprehensive patient data framework, providers can ask those questions directly. Natural language queries against the patient's chart, surfacing abnormal labs since the last visit, flagging documented changes in condition, and answering specific clinical questions. All of this becomes possible without leaving the workflow.
This is a meaningful shift in how providers interact with clinical information. Instead of navigating a chart and hoping they don't miss something, they can ask specific questions and get specific answers, with the underlying documentation available for review.
Connecting Pre-Visit Preparation to Documentation
Pre-visit summaries and ambient documentation aren't separate products with separate value propositions. They're two components of a continuous clinical workflow, one that starts before the patient arrives and ends with a complete note after they leave.
A provider who has reviewed an AI-generated pre-visit summary before the encounter walks in prepared, already thinking about the clinical priorities for this visit. When the encounter begins, and ambient documentation starts capturing the conversation, the provider is already operating at a higher level, asking better questions, engaging more directly with the patient's concerns, and making more informed decisions in real time.
The downstream effect is documentation quality. The clinical picture that the ambient AI captures is richer because the encounter itself was richer. Diagnoses are more specific. The assessment and plan reflect more precise clinical reasoning. Coding suggestions that emerge from the documentation are more accurate because the documentation more faithfully reflects the complexity of the care that was actually delivered.
What This Means for Providers
The conversation about AI in healthcare has spent a lot of time on what AI can do after the clinical encounter (documentation, coding, billing). But it represents only part of the opportunity.
The administrative and cognitive burden on physicians doesn't begin when the visit ends. It begins when the workday starts. Pre-charting is where providers spend time that they can't easily account for, that doesn't show up in RVU counts, and that no one has historically built technology to address.
Pre-visit summaries change that. They return time to providers at the front of the clinical day, not just the back. They reduce the cognitive load that accumulates before a provider has even greeted their first patient. And they create the conditions for a better clinical encounter, which is ultimately what both providers and patients are there for.
The technology to automate this work exists now. Providers who are already experiencing two hours of pajama time at the end of the day shouldn't also be spending significant time on manual pre-charting at the start of it. That's a problem with a solution, and AI is ready to deliver it.
Interested in learning more about how Commure's Patient360 platform and pre-visit summaries can work for your organization? Contact Dr. Jean-Luc Neptune, Clinical Commercial Leader at Commure, at jeanluc.neptune@commure.com.
What makes charge capture errors so dangerous is that they never show up as a denial. There's no rejection to appeal, no error message, no line on a dashboard. The charge simply never gets created, and by the time anyone notices, the timely filing windows have closed.
So before you tune your denials process or renegotiate a payer contract, look at the part of the revenue cycle where the leakage actually starts.
What charge capture is
Charge capture is the process of recording every billable service, supply, and procedure from a patient encounter so it can be coded and billed.
It happens at or right after the point of care, during charting, or immediately after the patient leaves. It's the handoff between the care a clinician delivers and the claim your billing team builds.
Two distinctions matter. Charge capture and coding are separate steps: coding translates the captured service into CPT and ICD-10 codes after the charge exists. Charge capture also depends on clinical documentation. If a service isn't documented, it usually won't be charged, meaning it won't be coded, billed, or paid.
Why revenue leaks at the charge
Charge capture breaks in quiet ways.
A clinician finishes a procedure and moves to the next patient before the charge is entered. An incision is made and an implant is placed in the OR, but no corresponding charge is recorded. An E&M visit gets documented at a lower complexity than the care actually delivered. Supplies used at the bedside never make it onto the bill.
Each miss is small. Across thousands of encounters a month, the small misses compound into real money.
Charge lag adds to the problem. The longer the gap between the service and the charge, the more detail gets forgotten, and the closer the claim drifts toward its filing deadline. Chargemaster mismatches make it worse: clinical practice changes faster than the codes and pricing logic behind it, so the charge that does get recorded may not reflect the full value of what was delivered.
The root cause is usually a manual process. When charge entry depends on memory, spreadsheets, and retrospective chart review, coverage is thin and slow. Most manual reconciliation touches a fraction of charts, and it happens after claims have already gone out.
To make things worse, these missed charges are permanent. You can rework a denied claim. You cannot bill for a service you never recorded once the filing deadline passes.
Hospital charge capture is harder at scale
The bigger the organization, the more places a charge can fall through.
Hospitals bill both facility and professional fees, and the two don't always line up. High-volume, high-acuity settings like the emergency department make real-time capture difficult because clinicians are focused on patients, not charge tickets.
Procedure-heavy specialties carry the most risk. Cardiology, orthopedics, radiology, and surgical services all involve dense documentation and high-dollar items that are easy to miss or undercount.
When charge capture depends on people remembering to act, every open seat on a revenue cycle team is a gap in coverage.
Eligibility, coding, claim submission, denials, and posting all depend on a charge existing in the first place. A clean claims rate can look healthy right up until you realize it can't recover a charge that was never created.
This is why charge capture deserves attention as a strategic control. It's the first point where clinical activity turns into revenue, and the one point where a loss can't be recovered.
How AI is changing charge capture software
For years, charge capture software meant a mobile app that let clinicians enter charges faster. Useful, but still dependent on a human remembering to enter the charge.
AI changes the model. Rather than waiting for someone to log a charge, the system reads the clinical note, compares what was documented against what was billed, and flags services that were documented but never charged.
This is the difference between charge capture and charge note reconciliation. Charge capture depends on someone remembering to act. Charge note reconciliation runs automatically on every encounter, whether or not anyone did.
That shift in coverage is the whole point. Manual review samples a slice of charts after the fact; automated reconciliation reads all of them before the claim goes out, which is when a missed charge can still be fixed.
Pair that with autonomous coding, which generates CPT and ICD-10 codes directly from the documentation, and the path from encounter to clean claim gets shorter and more accurate. The same approach turns charge capture solutions from a faster way to type into a safety net that runs automatically.
What this looks like in practice
The proof is in what happens when health systems close the gap.
HFMA documented Novant Health, a 14-hospital system with $4.3 billion in net patient revenue, that uncovered $7.5 million in recoverable revenue within 15 months of overhauling its charge capture process. That was revenue from services that had been delivered, documented, and then missed.
Commure built charge capture and charge note reconciliation into Commure Pro, the clinical intelligence platform clinicians use at the point of care, and connected it to the broader RCM platform so a captured charge flows straight toward the claim.
Charge capture is the front end of getting paid for care you've already delivered.
When it runs on memory and retrospective review, revenue leaks quietly and permanently. When the clinical note, the charge, and the code connect automatically, the leak closes and the revenue shows up where it should.
If you're rethinking where your revenue cycle loses money, start at the charge. It's the cheapest dollar to recover, because it's a dollar you already earned.
See how Commure connects charge capture to the rest of the revenue cycle.
Tell us a little bit about yourself—what do you like to do outside of work?
Outside of work, I spend most of my time with my family. I also love trying new restaurants and finding great local spots. When I finally get some downtime, you'll usually find me watching reality TV, which I consider an unofficial study in human behavior.
As a kid, what did you want to be when you grew up?
My dream was to play in the WNBA. Becoming an Operations Leader at a healthcare technology company was a close second. Unfortunately, I lacked both the height and basketball skills to make the first one happen.
Describe a day in the life of your role.
No two days are the same. My role is a mix of solving operational issues, improving processes, supporting clients, and partnering with teams across the organization to keep things moving. A typical day might include troubleshooting escalations, digging into data to identify trends, aligning with cross-functional teams on priorities, and removing blockers for both internal teams and clients. I usually start the day with a plan, but things shift quickly, so staying flexible and focused on impact is a big part of the job.
What made you decide to join Commure?
I was introduced to Commure by a former manager I had worked with previously. I respected their judgment and was intrigued by the opportunity to work on meaningful, complex problems in healthcare. When the role came up, it felt like the right next step. Two and a half years later, we're both still here, which says a lot about the people and the work.
How would you describe the Commure company culture?
.Gritty, driven, and fast-moving. People here care deeply about the mission and are willing to do what it takes to solve hard, often ambiguous problems. There is a strong bias toward action and ownership, and expectations are high in a good way. I've never worked with a group of people more committed to actually getting to the right outcome, not just talking about it.
What advice would you give someone on their first day at Commure?
Trust that you know more than you think you do. The learning curve can be steep, but one day you'll realize that what once felt chaotic now feels routine, and you'll recognize the role you played in making it better.
What has been your greatest accomplishment so far at Commure?
While I am proud of the operational improvements and results we have achieved, my greatest accomplishment has been the relationships I have built with my clients. Many of them have been with me since my very first day at Commure, and we have grown together through challenges, successes, and significant change. Looking back on how far we have come as partners, and knowing the trust we have built along the way, has been one of the most rewarding parts of my journey here. In many ways, their continued partnership is the accomplishment I am most proud of.
Interested in a career building the next generation of healthcare technology powered by AI? We are always looking for talented people across our departments.
Patients now shop for care the way they shop for everything else. They compare, they read reviews, and they leave when the experience frustrates them. A 2024 Deloitte survey of US health care consumers found that about 24% are willing to switch doctors just to get access to virtual care options.
That pressure has a name in healthcare strategy: the digital front door. And the market behind it is growing fast. Mordor Intelligence projects the digital front door market will expand from about $31.66 billion in 2026 to $82.25 billion by 2031.
For health system leaders, the question is how to build a digital front door that works. This guide covers what a digital front door is, the four components that make it up, why your patients expect one, and how to evaluate a platform.
What is a digital front door in healthcare?
A digital front door is the set of connected digital tools a health system uses to guide patients from their first interaction through scheduling, intake, communication, and ongoing care. It gives patients one consistent way to find care, book it, prepare for it, and stay connected after.
Think of it as the digital equivalent of walking up to the front desk, except it spans every step before, during, and after a visit. A patient might search for a provider, book online, complete intake on their phone, get reminders by text, and message the care team once they're home.
Done well, a digital front door connects these moments so patients don't have to repeat themselves or chase down information.
The four components of a digital front door
A digital front door is made up of four capabilities that work together. When health systems buy them as separate point solutions, they tend to create friction rather than remove it.
Patient scheduling and access
This is the most visible component, and often the first thing patients judge you on. Online self-scheduling lets patients book, reschedule, and cancel without calling a line that may sit on hold for several minutes. The advent of AI voice agents is also helping to ease the load on hospital staff while giving patients who prefer to speak to somebody an avenue to schedule care.
Scheduling sets the tone. A patient who can book in 90 seconds at 11 p.m. has a very different first impression than one stuck in a phone tree.
Patient intake
Once an appointment is booked, intake collects the information clinicians need before the patient arrives. Digital intake moves registration, insurance details, and clinical history off the clipboard and onto the patient's phone.
This is one of the highest-ROI components, because it cuts front-desk data entry and shortens waiting room time. We covered the shift in depth in our look at how AI is replacing the clipboard at patient intake.
Patient communication
Communication runs across the entire journey: appointment reminders, prep instructions, billing notices, and two-way messaging with the care team. Patients expect this on the channels they already use, mostly text and email.
Care navigation guides patients to the right next step, whether that's a specialist referral, a follow-up, or a post-discharge check-in. This is where a digital front door earns its keep over time.
Automated care pathways and reminders keep patients on track between visits, which is where most no-shows and care gaps happen. Our piece on AI-powered care journeys and no-show reduction walks through how this works in practice.
Why health systems need a digital front door
Three forces are pushing this from nice-to-have to expected.
First, patient expectations. Banking and retail have trained patients to expect 24/7 self-service, and they apply that standard to healthcare. Not to mention, when a patient is seeking care, they are likely in a vulnerable position and need help right away.
Second, competition for patients. Retail clinics, virtual-first providers, and large systems with mature digital tools are all chasing the same patients. A clunky access experience sends them elsewhere.
Third, operational pressure. Staffing shortages and rising costs mean phone and paper-based access doesn't scale. Digital tools (such as AI call center agents) absorb routine work so staff can focus on patients who need a human.
No-shows, abandoned appointment requests, and leakage to competitors all cost revenue. A working digital front door recovers some of that by making it easy for patients to start and stay in care with you.
How to evaluate a digital front door platform
Once you decide to invest, the build-versus-buy and single-platform-versus-point-solution questions matter more than any feature checklist. Four criteria separate platforms that work from those that create new silos.
EHR integration. The platform has to write back to your system of record. If scheduling, intake, and messaging data don't flow into the EHR, you've added work instead of removing it.
A unified platform. Patients feel it when scheduling, intake, and communication come from three different vendors. One platform covering all four components gives patients a consistent experience and gives you one place to manage it.
Omnichannel, compliant communication. Look for SMS, voice, and email in one place, with HIPAA-compliant messaging built in rather than bolted on.
Measurable outcomes. Ask vendors for proof: no-show reduction, response rates, staff time saved. Yale New Haven Health, for example, cut no-show and same-day cancellation rates by 54% in a breast imaging program built on this kind of outreach.
Frequently asked questions
Is a digital front door the same as a patient portal?
A patient portal is one piece of a digital front door, usually focused on records and messaging for patients who are already established. A digital front door is broader. It covers how new and returning patients find and book care in the first place.
What's the first component to prioritize?
Most systems start with scheduling and intake, since those touch every patient and show fast results in reduced phone volume and front-desk work. Communication and care navigation build on that foundation.
Who should own the digital front door in a health system?
Ownership usually sits across several teams, which is part of why these projects stall. Marketing cares about patient acquisition, operations cares about access and staffing, and IT owns the EHR. The systems that succeed name one owner or steering group with authority across all three, so the experience doesn't fragment by department.
Building your digital front door
Patients form an opinion about your health system before they ever meet a clinician. The digital front door is where that opinion starts, and it runs through every step that follows: how they book, how they check in, how they hear from you, and how they find their way to the next visit.
Commure Engage brings scheduling, intake, communication, and care navigation into one platform connected to your EHR, so patients get one front door instead of four. See how it fits your access strategy.
Every clinician knows the feeling. Packed schedule, patient about to walk in, and you're scrambling through charts trying to piece together context. All of this quietly contributes to burnout before the day has even started.
That pre-visit scramble is exactly what I set out to solve.
Patient 360 and the Pre-Visit Summary
At Commure, we've been building toward Patient 360 (a a knowledge engine that unifies all of a patient's data into a single clinical picture: active problems, current treatments, care history, and open follow-ups), and surfaces them at the moment they matter most.
Pre-Visit Summary is the first major feature built on top of Patient 360. A few weeks ago, we released it to over a thousand providers across multiple specialties. The response has been overwhelming.
A primary care physician put it simply: "Patient 360 gave me exactly what I needed to prep for the visit. The right level of patient-centered information surfaced at the right time. The reminders on incomplete tests were especially valuable."
And from an orthopedic surgeon, right before a busy clinic: "It gave me exactly the information I needed and saved me several minutes of going into the charts. Providers who normally wouldn't use AI tools are now gravitating towards this."
Built by Clinicians, for Clinicians
Leaning on my experience as a physician, I led the clinical strategy for this project.
Drawing on my own clinical judgment, I consulted physicians across specialties, validated the design against real workflows, and iterated with our Commure Clinical Council, Commure’s internal body of clinicians. All of this reinforced the central principle of our work: understanding a patient's key care considerations is foundational to establishing a therapeutic alliance. That's why the summary surfaces not only information about labs, medications, and other aspects of the current treatment, but also interactions with other providers and other significant life contexts. A provider walking into a room should know what else is going on in that patient's world.
Our goal for providers was simple: be able to glance through patient information in two to three minutes and walk in fully prepared, including what's still open, what needs follow-up, and where there are gaps that need reconciliation.
A Good AI Product Rests on a Good Data Strategy
We believe clinicians should drive data strategy because they know which information matters most and where it comes from. So physicians and clinicians at Commure led the work of defining a data hierarchy framework to determine what gets surfaced, in what order, and why.
That hierarchy had to solve for three things simultaneously: relevance (what an orthopedic surgeon needs differs from what a cardiologist needs), recency (information had to reflect the patient's current state), and longitudinal coherence (surfacing how a condition has evolved over time, not just a snapshot).
This wasn't a technical exercise with clinical input bolted on at the end. It was clinical thinking embedded into the architecture of the product itself.
This Is What Clinician-Led AI Looks Like
There's a version of AI in healthcare that automates tasks. And then there's a version that understands clinical workflow, anticipates what a provider needs, and surfaces it at exactly the right moment.
Pre-Visit Summary is the latter. But beyond reducing burden, what excites me about this project is its potential for patient impact. When providers are well-prepared, continuity of care improves. Nothing falls through the cracks and they can deliver better care while taking better care of themselves.
Anthropic released its first generally available Mythos-class model, Fable 5, a new tier of AI that sits above every previous frontier model. For the first time, the most advanced and capable class of AI models ever built was designed with clinical and operational healthcare work in mind.
Commure has done the security, compliance, and architecture work to deploy Mythos-class models within our healthcare-grade environment, and we now have the ability to apply them across the Commure platform.
What this class of model does for healthcare workflows
The defining trait of Mythos-class AI models is endurance. They sustain long, complex, tasks far beyond what previous generations could hold together, and they are measurably more reliable across the multi-step workflows healthcare operations run on.
Voice and call center agents. A patient call is a long-horizon reasoning problem: understand an unwell or frustrated caller, identify intent, navigate scheduling within the EHR, recover when the conversation jumps tracks, and know when a human should take over. Each underlying model generation improves the weakest link in that chain, which is sustained multi-turn reasoning under ambiguity. Mythos-class models are the largest single improvement to that link yet.
Patient intake. Intake agents gather history, reconcile it against the chart, follow clinically validated pathways, and hand the clinician a summary worth reading. High-fidelity synthesis of patient records and care histories is precisely where this model class excels, and the longer and messier the intake, the bigger its advantage.
Autonomous coding and revenue cycle. Producing correct CPT, ICD-10, and modifiers from documentation is dense, rules-heavy reasoning where small errors compound into denials. The frontier labs now consider this work important enough to benchmark: the newest evaluations in the model's own system card include prior authorization, denials and appeals, and DME orders, executed across simulated EHR and payer-portal environments. Cleaner coding upstream means fewer denials downstream and lower cost-to-collect.
Measured on healthcare work
Mythos-class models are state-of-the-art on nearly all tested benchmarks of AI capability, spanning software engineering, knowledge work, vision, and scientific research. Healthcare has rigorous benchmarks of its own, and the new model class was evaluated on them directly.
HealthBench Professional tests real provider workflows. It is built from 525 conversations authored by practicing physicians, spanning clinical consults, documentation, and research tasks, each graded against physician-written criteria. A model's score reflects whether its clinical reasoning and documentation would hold up in front of the clinician who wrote the task.
HealthBench tests how a model handles patients. It grades 5,000 multi-turn patient conversations across 26 medical specialties against more than 48,000 expert-written rubric items, measuring safety, accuracy, and communication in realistic healthcare contexts. This is the benchmark closest to what a patient-facing agent does all day.
In physician-authored clinical work, Fable 5 Mythos-class scores nine points above the previous flagship (Opus 4.8) and fourteen above GPT-5.5.
The Next Frontier of Agentic AI for Healthcare
These gains represent a fundamental shift in what agents can handle on their own, enabling more complex, multistep use cases that previously required human intervention at every turn. With Commure's deep healthcare expertise, data integration, and infrastructure, we can deploy these models in clinical settings at an even higher level of performance.
Apply the latest frontier models today
Mythos-class capability is now part of the Commure platform. If you are a current customer, you are already set as you keep the same deployment, workflows, and integrations. If you are not yet a customer, this is a good week to see what an LLM-native platform can do.
Tell us a little bit about yourself—what do you like to do outside of work?
I am a family man first and foremost. I've been married five years, have a son, and have another baby boy on the way. We recently relocated down to San Diego from the Bay Area and love it out here. Outside of spending time with family, I'm a huge sports enthusiast. There are not many days where you won’t see some sports event on the TV. Also, like staying active, golf is my game of choice and you'll find me on the course on weekends whenever I can sneak away.
As a kid, what did you want to be when you grew up?
A professional athlete, without a doubt. The dream was alive and well.
Describe a day in the life of your role.
No two days look the same, which is what I love about it. My role in Strategic Partnerships means I'm constantly connecting with external organizations, individuals, and clinics, identifying opportunities that can move the needle on our growth, and working with eng, ops, and other internal teams to bring those partnerships to life. It's part relationship-building, part strategy, and part execution/closing, and I genuinely enjoy all three.
What made you decide to join Commure?
I've spent the last 10 years in healthcare, starting on the med device side, and I reached a point where I wanted to feel like I had a direct hand in shaping the future of healthcare rather than just being another cog in the machine. Having family members in the medical field also made it personal. Knowing that the work we do here can make their lives easier and improve the system they work in every day and, in turn, provide better care to patients was a huge motivator for me. Joining Athelas/Commure gave me exactly that opportunity and it was too compelling to pass up.
How would you describe the Commure company culture?
Builders and meritocracy. Full stop. Everywhere you look, you see people coming in, building something real, and being rewarded for it. Career growth happens here at a pace you simply wouldn't find anywhere else. It pushes you to develop faster, think bigger, and take ownership in ways that permanently change how you work.
What advice would you give someone on their first day at Commure?
I have two, first, be a sponge. There are incredibly smart, talented, and successful people around you so learn from them, ask questions, observe, and absorb everything. Second, don't be shy about spotting areas for improvement. Fresh eyes are valuable, and if you see a way to make something better, speak up. That kind of initiative is noticed and welcomed here.
What has been your greatest accomplishment so far at Commure?
Being here from day one of selling RCM and growing into a role where I've managed teams across RCM and Scribe, and now working on strategic partnerships for our entire platform that are helping shape our GTM motion, is something I'm really proud of. It's been a full-circle journey, and the fact that I've had a hand in building something from the ground up at every stage makes it all the more meaningful.
Interested in a career building the next generation of healthcare technology powered by AI? We are always looking for talented people across our departments.
In healthcare, what gets documented gets paid for. What doesn't, doesn't.
That simple equation is why clinical documentation improvement (CDI, also known as clinical documentation integrity) has become one of the most consequential disciplines inside modern health systems. CDI sits at the intersection of clinical care, coding accuracy, regulatory compliance, and revenue. When it works, hospitals get paid appropriately for the care they deliver. Quality scores reflect reality. Clinicians spend less time chasing queries.
When it doesn't, the costs compound, resulting in denied claims, underpayments, distorted quality metrics, and frustrated providers.
The discipline isn't new. But the way leading health systems approach it is changing fast. The traditional CDI playbook was built around manual chart reviews and retrospective physician queries, designed for an era of paper records and fee-for-service billing. AI is rewriting that playbook.
Here's what CDI is, why it matters more than ever, and how AI is changing the way health systems approach documentation.
What is clinical documentation improvement?
Clinical documentation improvement is the practice of making sure a patient's medical record accurately reflects their clinical condition, the care they received, and the medical decision-making behind it. Done well, CDI bridges the gap between clinical reality and what eventually shows up on a claim form.
A traditional CDI program runs on certified specialists, usually nurses or coders with additional credentials. They review patient charts, identify gaps or ambiguities, and submit queries to physicians for clarification. They work alongside medical coders to make sure diagnoses are documented with the specificity required for accurate ICD-10 and CPT coding, that complications and comorbidities (CCs and MCCs) are captured, and that the final claim reflects the true acuity of the patient.
AHIMA's guidance for CDI programs frames the discipline as foundational to accurate reporting under CMS quality measures, value-based purchasing, and DRG-based payment.
Documentation determines whether a chart reflects the full acuity of a patient, and that distinction often drives thousands of dollars per case in reimbursement. A patient documented with a major complication or comorbidity (MCC) can generate more than double the payment of an otherwise identical chart without one. The clinical reality is the same. The documentation is what changes the number.
A peer-reviewed study in the Journal of Vascular Surgery found that a physician-led CDI initiative significantly increased the case mix index and contribution margin. That finding has been replicated across specialties and care settings.
This is also why CDI is foundational to revenue cycle management. Coding, claim submission, denials prevention, reimbursement: they all depend on documentation that meets payer and regulatory standards, with CDI sitting upstream of all of it.
Why clinical documentation improvement matters
CDI isn't a back-office function. Its impact shows up in three places every health system executive cares about: revenue, quality, and clinician experience.
Financial impact. Strong CDI programs have a direct line to the bottom line. AHIMA's Clinical Documentation Improvement Toolkit identifies CDI as one of the most effective levers a hospital has for reducing denials, capturing severity of illness and risk of mortality, and securing appropriate reimbursement. Health systems using modern CDI tooling embedded in EHR workflows have seen revenue gains of up to 3% when documentation gaps are flagged in real time. For an enterprise health system, 3 points of net patient revenue could be the difference between hitting and missing a year.
Clinical quality and risk adjustment. Documentation determines how the outside world sees a hospital's quality. Mortality indices, publicly reported outcomes, case mix index (CMI), and risk-adjusted payment under value-based contracts all flow from what gets documented. CMS continues to expand value-based care arrangements, which makes precise documentation a revenue, quality, and contractual question all at once.
The old CDI playbook is running out of road for one fundamental reason. It relies on humans reviewing charts after the fact, and there aren't enough humans to do it.
Layering more retrospective queries onto already-burned-out physicians is a tax. And it's a tax that traditional CDI programs depend on to function.
How AI is reshaping clinical documentation improvement
The real shift in CDI isn't faster query workflows or better dashboards. It's that AI moves documentation improvement from a retrospective discipline to a real-time one.
AI-assisted CDI expands specialist capacity. According to MedLearn Publishing's analysis of AI in clinical documentation integrity, CDI specialists augmented by intelligent systems can manage 35–45% larger chart volumes, which lets organizations expand CDI coverage without proportional staffing increases. The same analysis reports 12–18% denial reductions for complex inpatient stays and 20–25% reductions on prior authorization submissions when AI-assisted CDI is deployed at scale.
These gains reflect a different operating model for CDI, one where documentation improvement is built into the workflow itself instead of bolted on afterward.
What to look for in a modern CDI approach
For health system leaders evaluating where CDI is headed, four questions matter:
Is documentation improved in real time or retrospectively? Real-time intervention prevents errors. Retrospective intervention chases them.
Is the system EHR-integrated or bolted on? CDI that requires clinicians to switch contexts or duplicate work won't get adopted.
Does it connect documentation to coding to claims? The biggest gains come from treating the documentation-to-reimbursement pipeline as one continuous workflow instead of three disconnected ones.
Is it built for enterprise scale? Pilots are easy. Sustained CDI improvement across hundreds of providers and dozens of specialties is the real test.
.avif)
.gif)
%20(2).gif)
.jpeg)
