The Numbers Behind the Hours
The American Medical Association's annual Physician Practice Benchmark Survey has tracked documentation time for more than a decade. The 2023 edition found that outpatient physicians — across specialties — spend a median of 2.1 hours daily on EHR documentation and administrative tasks directly tied to the clinical note. For a standard 220-day working year, that is 462 hours. Divided by an eight-hour workday, that is roughly 57.75 working days per year.
Not 26. We said 26. The discrepancy matters, and it tells you something about how the framing shapes the urgency differently for different audiences.
If you count only after-hours documentation — the chart work that spills into evenings and weekends after the last patient leaves — the median figure across AMA data is closer to 1.2 to 1.4 hours per day. Multiply that over a year, and you arrive at approximately 26 to 30 extra-hours working days: time that belongs to the physician's personal life, now occupied by the EHR. That is the number that tends to land with physicians when you say it out loud. It is not hyperbole. It is arithmetic.
Why the EHR Didn't Fix the Problem It Was Supposed to Solve
The original pitch for electronic health records — especially as HITECH Act incentives accelerated adoption after 2009 — included better care coordination, fewer medication errors, cleaner billing, and, implicitly, a more efficient clinical workflow. By most measures, the first three improved. The fourth did not.
What happened is now well-documented in health services research: EHR systems were built around billing and regulatory compliance requirements first, and clinical workflow second. Point-and-click documentation, pull-down menus, and templated fields made charts structurally consistent for coding purposes while making them cognitively taxing to complete and genuinely difficult to read. The phenomenon researchers call "note bloat" — where a meaningful encounter generates five pages of auto-populated boilerplate — is a direct consequence of this architecture.
The result is that physicians spend significant cognitive effort translating the actual clinical encounter — a human conversation with diagnostic reasoning threaded through it — into a format the EHR was designed to accept. That translation work is the bottleneck. It is not that physicians type slowly or are unfamiliar with their EHR system after years of use. It is that the conversion from clinical thinking to structured record is inherently effortful and time-consuming when done manually.
What the Time Actually Costs
The burnout literature is consistent: documentation burden is among the top two or three cited contributors to physician burnout. The Medscape Physician Burnout Report, which surveys tens of thousands of physicians annually, has listed "too many bureaucratic tasks" as the leading burnout driver for six consecutive survey years. Documentation is the largest single component of that category.
The downstream costs are not just personal. When a physician is burned out, patient-facing time and quality of presence suffer. Decisions made in a state of cognitive overload carry different risk profiles than decisions made by a physician whose working memory is not occupied by the note they still need to file. Some of the most important patient interactions happen in that final minute of an encounter — the moment when a patient finally asks the question they came in to ask. A physician who is mentally drafting the Assessment and Plan section while the patient speaks is less present for that moment.
At an institutional level, burnout-driven attrition is expensive. Estimates for the cost of replacing a physician — recruiting, credentialing, lost revenue during the gap — range from $500,000 to over $1 million depending on specialty. If documentation burden is a meaningful contributor to departure decisions, the ROI calculus for tools that reduce it becomes straightforward for a CFO or CMIO.
The Cascade Internal Medicine Example
Consider a scenario representative of what growing independent practices encounter. A 12-physician internal medicine group — call them Cascade Internal Medicine, a composite of patterns we see across similar practices — finds that physicians are averaging 90 minutes of after-visit documentation per day. Over a year, that is roughly 330 hours per physician, or 3,960 physician-hours across the group. At a conservative loaded cost of $150 per physician hour, that is nearly $600,000 in labor allocated to chart completion rather than patient care or rest. None of that time generates additional revenue. Most of it generates resentment.
Ambient AI Is a Different Kind of Intervention
Tools designed to reduce documentation time have existed in various forms for decades. Early medical transcription services — where physicians dictated and a human transcriptionist produced the note — reduced the typing burden but added latency (often 24-48 hours for note completion), cost, and a step in the chain where clinical meaning could be miscaptured. Dragon-style voice recognition eliminated the latency but required physicians to essentially narrate a complete note, which many found to be a different kind of cognitive work rather than a genuine time savings.
Ambient clinical AI takes a meaningfully different approach. Rather than requiring the physician to produce the note verbally, the system listens to the clinical encounter as it happens — the actual conversation between physician and patient — and builds a structured SOAP note from that conversation. The physician does not need to narrate documentation mode. They speak to their patient. The note emerges from that interaction.
We are not saying ambient AI is a complete solution to physician burnout. Documentation burden is one factor among several — prior authorizations, inbox management, message volume, and care coordination overhead are also substantial. Addressing the documentation bottleneck is meaningful but not sufficient on its own.
What ambient AI does address is the specific, time-intensive translation step between clinical encounter and written record. When that step is automated, the physician's role shifts from author to reviewer. And reviewing a structured draft note — scanning it for accuracy, correcting any errors, adding clinical nuance the system missed, and signing — typically takes 60 to 90 seconds rather than 15 to 20 minutes. That is the order of magnitude that moves the needle on the 26-day calculation.
What Accurate Ambient AI Requires
Not all ambient documentation tools perform equivalently. The technical problem is significantly harder than most vendors acknowledge. Clinical speech is not general speech: it contains specialty-specific terminology, abbreviations, dosing conventions, and diagnostic reasoning patterns that a general speech-to-text model handles poorly. A system that confuses "15 milligrams" with "50 milligrams," or that captures a physician's hypothetical ("if this were a pulmonary embolism, we'd expect...") as a diagnostic statement, creates more work for the physician reviewer than it saves.
The ICD-10 mapping layer adds another constraint. Generating a plausible-sounding Assessment section is easier than generating the correct ICD-10 code to the fourth or fifth character level of specificity. A note that lists "type 2 diabetes mellitus" in the Assessment without specifying whether there are complications, and if so which, will produce a less specific code than the encounter actually supports — with downstream implications for reimbursement and quality reporting.
Good ambient AI requires a clinical NLU layer purpose-built for the documentation task: one that understands symptom-diagnosis mapping, medication reconciliation logic, and the difference between a physician's exploratory reasoning and their clinical conclusion. This is not a solved problem universally — it is an active area of differentiation among tools in this space.
Getting Back to 26 Days
The 26-day framing is worth sitting with for a moment. Most physicians, when they hear it, recognize the number immediately. Not because they have calculated it, but because they feel it. The evenings at the computer after the kids are in bed. The notes flagged for completion over the weekend. The sense that the job never fully ends because the documentation clock keeps running even when clinical hours are over.
Addressing that specifically — not burnout in the abstract, not moral injury at a systemic level, but the concrete 26 days — is a tractable problem. The underlying technology has reached a maturity point where practices of almost any size can evaluate ambient documentation tools with realistic expectations about what they will and won't do. The evidence for time savings is real. The evidence for note quality maintenance (or improvement) is growing.
The question for a physician considering this category of tool is not whether the technology works. It is whether the specific tool works for their encounter types, their EHR, and their clinical style — and whether the review workflow it creates actually feels faster than what they do today. That is an empirical question, answerable in a 30-day pilot. Most physicians who do that evaluation reach a clear conclusion.