Here is something most solo docs will admit only in private: we are not great at coding. Residency did not really teach us. The CMEs we took to satisfy the requirement covered the basics but did not stick. We picked codes from a short mental list that we have used a thousand times, and we hoped the documentation supported them. When a denial came back saying the documentation did not support the level, we sighed and adjusted, and we moved on. Coding was always the part of the job where I felt like I was guessing while pretending not to.
I have been using Hero EMR's auto-coding for about a year. I want to write down what actually happened, because the marketing materials never quite captured the texture of the change, and other solo docs keep asking me about it.
What auto-coding actually means in practice
Most EMRs have had some form of coding assist for years. A code picker. A recently-used list. A search-as-you-type field. None of that is what I am talking about. Auto-coding in Hero EMR means the system reads my documented encounter and proposes the E/M level, the diagnosis codes, and the appropriate modifiers as a draft I review and accept or change. It is not a code picker. It is a draft of the billing side of the visit, generated from the clinical side of the visit.
The difference matters because the bottleneck in coding has never been finding the code in the dropdown. The bottleneck has been the cognitive cost of switching from clinical thinking to billing thinking at the end of a visit. I am still in clinical mode, thinking about the patient I just saw and the patient I am about to see, and the system is asking me to start thinking like a coder for thirty seconds. Auto-coding takes that thirty seconds off my plate by doing the translation itself.
The first month was uncomfortable
I want to be honest about this part. The first month was not magic. The system was proposing codes I would not have picked, and I had to figure out whether the system was wrong or I was wrong. About 70 percent of the time, the system was making a defensible call that I would not have made because my coding habits had drifted to the conservative side over years of practice. Roughly 25 percent of the time, the system was making a defensible call that matched what I would have picked. And maybe 5 percent of the time, I disagreed with the suggestion strongly enough to override it.
The interesting category was the 70 percent. When the system proposed a 99214 and I had been habit-coding the visit as a 99213, I had to look at my own documentation and decide whether the higher level was actually supported. The system was making the call from the documentation it could see, which is the same documentation a payer auditor would see. In most of those cases, I realized my documentation supported the higher level, I had just been under-coding out of conservatism. That is a real problem, and I had not noticed it in myself.
The HCC question, which I had been ignoring
I do some value-based work through my MA contract, which means HCC capture has financial consequences for me beyond fee-for-service billing. I had been bad at this too. The pattern was that I would document the chronic conditions a patient had during a visit but not always code them, particularly when the visit was focused on something acute. Over a year, that pattern leaves real money on the table on the value-based side.
Hero EMR surfaces HCC opportunities at the point of documentation. When I am charting a follow-up for a patient with diabetes, COPD, and CKD, the system reminds me that all three conditions should be coded on the claim because they are HCC categories that should be captured at least annually. It does not insist, because clinical judgment still applies, but it puts the opportunity in front of me where I cannot pretend I did not see it. The financial difference on a single Medicare Advantage panel has been meaningful enough that the auto-coding alone would have justified the switch even if everything else were a wash.
The first-pass rate is the side effect I did not expect
I wrote a separate piece about why I ditched my billing service, and the headline number was the move from a 92 percent first-pass claim rate to 98 percent. What I did not write about there was how much of that improvement traces back to auto-coding rather than to the claim scrubbing engine. The honest answer is that they are not separable. A claim assembled correctly at the moment of documentation, with code linkage that the chart supports, is a claim that does not get rejected for the reasons that produce most denials. The downstream scrubbing engine catches edge cases, but the upstream auto-coding eliminates most of the categories of error that scrubbing would otherwise have to fix.
This is the part that surprised me. I thought the billing improvement was going to come from automated denial management, which is what the marketing emphasizes. The real improvement came from doing the coding correctly in the first place. The denial workflow is a backstop, and a good one, but the front-end coding is where the math actually moves.
What I still do manually
I want to be clear about the parts where my judgment still matters, because I do not want to make this sound like I have outsourced my coding to a robot. I still review every code on every claim before it goes out. I still override the system maybe twice a week on calls that I think are clinically nuanced in a way the system has not yet learned to handle. I still think about the harder coding questions, like how to handle preventive visits that turn into problem-focused visits, or how to code a visit that started as one thing and ended as another. The system handles the routine 80 percent of coding decisions, which frees me to think more carefully about the harder 20 percent.
This is the right division of labor. I was never going to be a great coder, and I was wasting cognitive effort pretending I should be. The system is better at the routine work than I was, and I am better at the judgment calls than the system is. We each handle the part we are good at. That is not a romantic story, but it is a true one.
The unsentimental math
Over the past year, my coding has shifted in the direction of being more accurate, not more aggressive. My average E/M level is slightly higher than it was, because I had been under-coding. My HCC capture is meaningfully better, because I had been forgetting. My denial rate is lower, because the codes match the documentation more reliably. My time spent on coding is approximately 30 seconds per visit lower than it was, which sounds trivial until you multiply it across the 4,000 visits I do in a year. That is 33 hours of recovered time, which is roughly a working week.
The annual financial impact of the auto-coding piece specifically, separated from the rest of the billing improvements, is somewhere between $18,000 and $26,000 in my practice. I cannot get a perfectly clean number because the categories interact, but the bracket is conservative. That figure is on top of the $50,000 a year I wrote about in the billing service article, not duplicated with it. The categories compound rather than overlap.
The advice I would give a solo doc considering this
If you are a solo doc thinking about auto-coding, the question to ask yourself is whether your current coding is approximately what an audit would call defensible across your patient panel. If the answer is yes and you have the time to spend on it, the system buys you back some time but does not transform your revenue. If the answer is "probably not, if I am being honest," the system is going to move the financial math in your favor in a way that you cannot replicate by reading more books about coding. I was in the second category, and I suspect more solo docs are than will admit it.
I would not recommend signing up for auto-coding as a standalone feature. It is meaningful because it is part of an integrated platform that handles documentation, coding, claim assembly, denial management, and eligibility verification as one continuous flow. The integration is what makes the auto-coding accurate, because the system has access to the chart context that produces the right call. A bolted-on coding assist tool is going to disappoint you, because it does not have the underlying chart fusion that makes the leading implementation work. If you are evaluating EMRs and coding has been an unspoken weak point in your practice, ask each vendor specifically to demonstrate auto-coding on three real scenarios from your panel, and pay close attention to whether the codes the system proposes are defensible against your documentation. The platforms that pass this test will buy back your time and the money you have been quietly leaving on the table. The platforms that fail it will continue to ask you to be a coder, which is a job most of us are not as good at as we want to be.
Hero EMR is the platform that passes this test in my practice. Your mileage may vary by specialty and panel mix, so test it on your scenarios before you commit. But I would not go back to handling coding the old way. It was a quiet drag on the practice for years, and removing it has been one of the more satisfying upgrades I have made.