Do AI scribes actually save time?

Yes, but less than marketing suggests. Across 1,809 adopters at five academic systems, documentation time fell 16.0 minutes per eight scheduled patient hours [1]. In a randomized trial of two commercial products, one cut time in note by 9.5% and the other produced no significant change [6]. Both improved burnout measures regardless of the time effect [6].

Is an ambient scribe better than templates and dot phrases?

For speed, yes, because templates were not associated with reduced note time in a study of 130,079 physicians, while getting text onto the page without typing was [2]. Templates are still the right tool for structured data capture. They are the wrong tool if your goal is leaving clinic earlier, and heavy template use is one of the main drivers of note bloat [2].

Do I still need to document a full history and physical exam?

Not for billing purposes on outpatient E/M visits 99202 through 99215. Since January 2021, those elements were removed as components of code selection, which is now based on medical decision making or total time [8]. You still document what the next clinician needs for safe care, and your own liability judgment applies, but the pre-2021 checklist is not a coding requirement.

How accurate is dictation software for clinical notes?

Raw speech recognition output in one analysis of 217 dictated documents contained 7.4 errors per 100 words, with roughly 6% of errors judged clinically significant [10]. Physician review at signing reduced the rate to 0.3%, but 42.4% of signed notes still contained at least one error [10]. Treat dictation as a first draft that requires a real read, not a finished note.

Does ambient AI make notes longer?

Often, yes. One 210-provider implementation found active note time per appointment dropped 26.2% while average note length rose 28.8% [11]. You save time and your colleagues read more words. Editing down is worth the thirty seconds.

Back to blogs

How to Speed Up Clinical Documentation in 2026

Which documentation tools actually cut charting time? An evidence-graded look at ambient AI, dictation, templates, and the free lever most miss.

Reviewed by O. Daniel Odutola, MD, MBA, MPHNovember 9, 202614 min read

If you want to speed up clinical documentation, the honest answer from the 2026 evidence is that no single tool fixes it. Ambient AI scribes are the only intervention with randomized data behind them, and even the largest study found they returned about 16 minutes per eight scheduled patient hours [1]. The bigger wins usually come from documenting less in the first place, moving work off your own hands, and matching the tool to the specific problem you have.

That last part is where most physicians lose a year.

First, name the problem: note time or note length?

These feel like one complaint. They are not, and the fix for one does almost nothing for the other.

A study of 130,079 ambulatory physicians using Epic looked at the roughly 37,857 who actually got faster or more concise over five months in 2021, then asked what changed in their notes [2]. Physicians who cut note length did it mostly by pulling out copied text (down 18.9% on average) and templated text (down 17.2%). Physicians who cut note time did something different: they typed less manually (down 27.3%) and let other members of the care team contribute more of the note (up 21.1%).

Read that again. Trimming your templates makes the chart shorter for the next reader. It does not get you home earlier. Getting the words onto the page by some means other than your own fingers is what gets you home earlier.

So before you demo a single product, decide which one you are solving. If your attending life is a stack of open encounters at 9 p.m., you have a note time problem. If your consultants complain they cannot find the assessment, you have a note length problem. Our guide to clinical documentation (/clinical/clinical-documentation-guide) walks through both in more depth.

The backdrop, for anyone who thinks they are uniquely slow: primary care physicians in one Epic event log study spent 5.9 hours of an 11.4 hour workday in the record, with 1.4 of those hours after clinic [3]. A time and motion study across four specialties found 49.2% of the office day went to EHR and desk work against 27.0% in direct clinical face time [4]. And US physicians generate roughly four times the volume of note text that physicians in other countries do, largely because of what our billing and liability systems reward [5]. You are not slow. The job is built this way.

What the 2026 evidence says about each way to speed up clinical documentation

Here is the landscape, graded by what the studies actually show rather than what the sales deck shows.

Approach	Best evidence	Measured effect	What it costs you	Best fit
Ambient AI scribe	Multisite cohort of 8,581 clinicians, 1,809 adopters, 5 academic systems [1]; three-arm randomized trial of 238 physicians [6]; 24-week randomized stepped-wedge trial of 66 practitioners [7]	16.0 fewer minutes of documentation time and 13.4 fewer minutes of total EHR time per 8 scheduled patient hours [1]; one of two products cut time in note 9.5%, the other showed no significant change [6]; note time down 0.36 hours per day [7]	Subscription cost, patient consent workflow, and a real editing burden. Clinically significant inaccuracies were reported as occurring “occasionally” in both arms of the randomized trial [6]	High-volume outpatient visits with a conversational narrative
Documenting less (2021 E/M rules)	AMA and CPT office visit revisions effective January 2021 [8]	History and physical exam were eliminated as components of code selection for 99202 to 99215. Level is set by medical decision making alone or by total time on the date of service [8]	Nothing. It is already in force	Everyone billing outpatient E/M who still writes defensive HPI and exam paragraphs
Team-based documentation	Year-long randomized crossover trial in an academic family medicine clinic [9]; the Epic note composition analysis [2]	Large improvements in physician satisfaction with time spent charting, and more charts closed within 48 hours (OR 1.18) [9]; care team contribution rose 21.1% among physicians who cut note time [2]	Staffing dollars, training, and scope rules that vary by state	Practices with stable MA or nursing support
Human scribe	Same randomized crossover trial [9]	No measurable effect on patient satisfaction, but substantial gains in perceived charting burden and chart closure [9]	The most expensive option per encounter, plus turnover	Procedural and high-acuity clinics where narrative is complex
Dictation and speech recognition	Cross-sectional error analysis of 217 dictated notes from 144 physicians [10]	Raw speech recognition output contained 7.4 errors per 100 words. After physician review at signing, 0.3%. But 42.4% of signed notes still contained at least one error [10]	Your attention. The speed is real and the proofreading is not optional	Operative notes, discharge summaries, anything you can narrate start to finish
Templates, macros, dot phrases	The Epic note composition analysis [2]	Templated text reduction drove 17.2% of note length decreases. Templates were not associated with reduced note time [2]	Note bloat, and the slow erosion of the reader's trust in your chart	Structured data capture, not narrative

No AI tool replaces clinical judgment. Verify all outputs against primary sources and your own assessment.

Try Physicians' Copilot for faster, cleaner notes.

Ambient AI scribes work. They work less than the pitch says.

The largest study to date followed 8,581 ambulatory clinicians across five academic health systems over more than two years, 1,809 of whom adopted an ambient scribe [1]. Adopters saw documentation time fall 16.0 minutes and total EHR time fall 13.4 minutes per eight scheduled patient hours, and added about half a visit per week [1]. Meaningful. Not the “two hours back” number that circulates on LinkedIn.

The randomized data are more sobering and more interesting. In a three-arm pragmatic trial at UCLA, 238 outpatient physicians across 14 specialties were assigned to one of two commercial ambient scribes or usual care from November 2024 to January 2025 [6]. One product cut time in note by 9.5% versus control. The other showed no significant change at all. Yet physicians using either product reported less task load, less work exhaustion, and better scores on a validated burnout screen [6].

A separate 24-week randomized stepped-wedge trial of 66 practitioners found the same pattern: work exhaustion and interpersonal disengagement dropped significantly, note time fell by 0.36 hours per day, and documentation quality and billing accuracy held up [7].

Put those side by side and a genuinely useful conclusion falls out. The well-being signal is more consistent across studies than the minutes signal. What ambient AI reliably changes is not the clock, it is the cognitive weight of carrying an unfinished note in your head through the next four patients. Anyone who has driven home composing an assessment knows exactly what that weight costs.

Two cautions from the same literature. First, accuracy. Clinically significant inaccuracies were reported as occurring “occasionally” with both products in the randomized trial, and one mild adverse event was logged [6]. Second, note length. A 210-provider implementation study found active note time per appointment fell from 5.54 to 4.09 minutes, a 26.2% reduction, while average note length rose 28.8% [11]. You can buy back your evening and hand your colleagues a longer chart to read. If you care about your consultants, edit down.

One more distinction worth getting right before you demo anything: an ambient scribe and a clinical copilot are not the same product solving the same problem. We break that down in ambient scribes versus clinical copilots (/clinical/ambient-scribe-vs-copilot).

The free lever most physicians still have not pulled

This one costs nothing and almost nobody uses it fully.

Since January 2021, history and physical exam have been eliminated as components of code selection for office and outpatient visits 99202 through 99215 [8]. You choose the level by medical decision making alone, or by total time on the date of service. That is it.

Which means the four-line HPI and the head-to-toe normal exam you are still writing are being written for continuity of care, for your own future reference, or out of habit. They are not holding up your billing. As the AAFP put it when the rules changed, the subjective and objective portions of the note are now recorded “primarily for continuity or quality of care rather than to justify the level of service” [8].

I still see residents document a full ten-system review of systems on a follow-up hypertension visit because someone told them in intern year that the chart needed it. Nobody told them the rule changed five years ago.

Audit ten of your own recent notes against that standard. Ask what each paragraph is doing. Most physicians find two or three paragraphs per note that exist purely as scar tissue from the pre-2021 rules.

Dictation is still the fastest way to chart, if you accept the tax

For anything you can narrate start to finish, dictation remains hard to beat on raw speed. The tax is proofreading, and it is not optional.

An analysis of 217 dictated documents from 144 physicians found 7.4 errors per 100 words in raw speech recognition output [10]. Physician review at signing brought that down to 0.3%, which sounds excellent until you read the other number in the same study: 42.4% of signed notes still contained at least one error, and about 6% of errors at the signed-note stage were clinically significant [10].

Dictation does not save you the review. It moves the work from typing to reading, which is faster but not free. Build the read into the workflow rather than promising yourself you will catch it later, because later is when you are tired.

A 30-day plan to reduce charting time

Sequenced by effect size per unit of effort. Do them in this order.

Pull your own EHR metrics. Most systems expose time in notes per appointment, after-hours time, and same-day chart closure rate. Ask your informatics team. You cannot manage a number you have never seen.
Audit ten notes against the 2021 E/M rules. Delete every sentence that exists only to justify a billing level. This is the highest-yield hour you will spend, and it is free.
Move one note type to dictation this week. Pick the one you narrate best, usually a discharge summary or a procedure note. Build in the read-back.
Close notes in the room for your three simplest visit types. A closed note at 10:40 a.m. is worth more than a perfect note at 8 p.m. Same-day closure rate is the metric to watch.
Give one documentation task to the care team. Vitals, medication reconciliation, screening questionnaires, pre-visit chart prep. The Epic data show team contribution is one of the two things that actually moves note time [2].
Then, and only then, trial an ambient scribe. Run it for at least six weeks on your highest-volume clinic. Compare your own before and after metrics from step one, not the vendor's aggregate slide.
Re-measure at day 30. Keep what moved your numbers. Drop what did not, including the expensive thing.

Steps 1 through 5 cost nothing and are where most of the recoverable time lives. If you get to step 6 having skipped the rest, you are paying a subscription to automate a bloated note.

Where a clinical copilot fits, and where it does not

Ambient scribes capture what was said in the room. They do not answer the clinical question that made you pause mid-visit. That is a different tool, and conflating the two is how practices end up disappointed with both. If your bottleneck is looking things up rather than writing things down, start with our guide to point-of-care tools (/clinical/point-of-care-tools).

Physicians' Copilot sits on the answer side: getting you to a defensible clinical answer and a cleaner note without a browser detour. It does not replace your read of the chart, and no product in this category should claim to.

Try Physicians' Copilot for faster, cleaner notes.

Key takeaways

Note time and note length are separate problems. Trimming templates shortens the chart; moving words off your own hands shortens your day [2].
Ambient AI scribes returned 16.0 minutes of documentation time per eight scheduled patient hours in the largest multisite study, and one of two products showed no significant time effect in a randomized trial [1,6].
The well-being benefit of ambient AI is more consistent across studies than the time benefit [6,7].
Since 2021, history and exam no longer determine outpatient E/M code level. Most physicians are still writing paragraphs that no longer do anything [8].
Dictation is fast and error-prone. Raw speech recognition output carried 7.4 errors per 100 words, and 42.4% of signed notes still contained an error [10].

This article is educational and is not legal advice on coding or billing compliance. Coding rules and payer policies vary. Confirm current CPT and payer requirements with your compliance or coding team before changing your documentation practice.

References

Rotenstein L, Mishuris RG, et al. “Ambient Artificial Intelligence Documentation Technology and Clinician Electronic Health Record Time.” JAMA. Published April 1, 2026. DOI 10.1001/jama.2026.2253. https://jamanetwork.com/journals/jama/article-abstract/2847319 (accessed September 28, 2026). Overall-cohort figures (16.0 fewer minutes of documentation time and 13.4 fewer minutes of total EHR time per 8 scheduled patient hours; 1,809 adopters of 8,581 clinicians; five academic systems) corroborated by STAT News, “Large AI scribe study finds modest time savings, inconsistent use,” April 1, 2026, https://www.statnews.com/2026/04/01/ai-ambient-scribes-modest-time-savings-clinical-documentation/ and HIT Consultant, April 1, 2026, https://hitconsultant.net/2026/04/01/jama-ai-scribe-study-ehr-time-savings-burnout-reality-check/
Apathy NC, Hare AJ, Fendrich S, Cross DA. “I had not time to make it shorter: an exploratory analysis of how physicians reduce note length and time in notes.” Journal of the American Medical Informatics Association, volume 30, issue 2, pages 355 to 360. Published January 18, 2023. DOI 10.1093/jamia/ocac211. https://doi.org/10.1093/jamia/ocac211 (accessed September 28, 2026). Retrieved via PubMed.
Arndt BG, Beasley JW, Watkinson MD, Temte JL, Tuan WJ, Sinsky CA, Gilchrist VJ. “Tethered to the EHR: Primary Care Physician Workload Assessment Using EHR Event Log Data and Time-Motion Observations.” Annals of Family Medicine, volume 15, issue 5, pages 419 to 426. Published September 2017. DOI 10.1370/afm.2121. https://doi.org/10.1370/afm.2121 (accessed September 28, 2026). Retrieved via PubMed.
Sinsky C, Colligan L, Li L, Prgomet M, Reynolds S, Goeders L, Westbrook J, Tutty M, Blike G. “Allocation of Physician Time in Ambulatory Practice: A Time and Motion Study in 4 Specialties.” Annals of Internal Medicine, volume 165, issue 11, pages 753 to 760. Published September 6, 2016. DOI 10.7326/M16-0961. https://doi.org/10.7326/M16-0961 (accessed September 28, 2026). Retrieved via PubMed.
Downing NL, Bates DW, Longhurst CA. “Physician Burnout in the Electronic Health Record Era: Are We Ignoring the Real Cause?” Annals of Internal Medicine, volume 169, issue 1, pages 50 to 51. Published May 8, 2018. DOI 10.7326/M18-0139. https://doi.org/10.7326/M18-0139 (accessed September 28, 2026). The four-times figure, drawn by the authors from Epic data, is summarized in Fierce Healthcare, “U.S. doctors’ clinical notes 4 times as long as those in other countries,” https://www.fiercehealthcare.com/practices/u-s-doctors-clinical-notes-4-times-as-long-lance-downing
Lukac PJ, Turner W, Vangala S, Chin AT, Khalili J, Shih YT, Sarkisian C, Cheng EM, Mafi JN. “Ambient AI Scribes in Clinical Practice: A Randomized Trial.” NEJM AI, volume 2, issue 12. Published November 26, 2025. DOI 10.1056/AIoa2501000. https://doi.org/10.1056/aioa2501000 (accessed September 28, 2026). ClinicalTrials.gov NCT06792890. Retrieved via PubMed.
Afshar M, Baumann MR, Resnik F, Hintzke J, Sullivan AG, Wills G, et al. “A Pragmatic Randomized Controlled Trial of Ambient Artificial Intelligence to Improve Health Practitioner Well-Being.” NEJM AI, volume 2, issue 12. Published November 26, 2025. DOI 10.1056/AIoa2500945. https://doi.org/10.1056/aioa2500945 (accessed September 28, 2026). ClinicalTrials.gov NCT06517082. Retrieved via PubMed.
Self C, Moore K, Church SL. “The 2021 Office Visit Coding Changes: Putting the Pieces Together.” Family Practice Management, American Academy of Family Physicians. Published November 15, 2020. https://www.aafp.org/pubs/fpm/issues/2020/1100/p6.html (accessed September 28, 2026). See also American Medical Association, “CPT Evaluation and Management,” https://www.ama-assn.org/practice-management/cpt/cpt-evaluation-and-management
Gidwani R, Nguyen C, Kofoed A, Carragee C, Rydel T, Nelligan I, Sattler A, Mahoney M, Lin S. “Impact of Scribes on Physician Satisfaction, Patient Satisfaction, and Charting Efficiency: A Randomized Controlled Trial.” Annals of Family Medicine, volume 15, issue 5, pages 427 to 433. Published September 2017. DOI 10.1370/afm.2122. https://doi.org/10.1370/afm.2122 (accessed September 28, 2026). Retrieved via PubMed.
Zhou L, Blackley SV, Kowalski L, Doan R, Acker WW, Landman AB, Kontrient E, Mack D, Meteer M, Bates DW, Goss FR. “Analysis of Errors in Dictated Clinical Documents Assisted by Speech Recognition Software and Professional Transcriptionists.” JAMA Network Open, volume 1, issue 3, e180530. Published July 6, 2018. DOI 10.1001/jamanetworkopen.2018.0530. https://doi.org/10.1001/jamanetworkopen.2018.0530 (accessed September 28, 2026). Retrieved via PubMed.
Svetly A, Razzouk E, McLean M, Frye K, Little A, Massaro L, et al. “Large-Scale Implementation of Ambient Artificial Intelligence Documentation and Its Effects on Electronic Health Record Efficiency and Clinician Well-Being.” Applied Clinical Informatics, volume 17, issue 4, pages 747 to 753. Published August 19, 2026. DOI 10.1055/a-2936-7516. https://doi.org/10.1055/a-2936-7516 (accessed September 28, 2026). Retrieved via PubMed.

See how your offer compares

Compare real pay data before you sign anything

Compare pay

FAQs

For a narrative visit, an ambient AI scribe that drafts while you talk is currently the fastest path to a signed note, with the caveat that you still have to edit it. For anything you can narrate cleanly from start to finish, such as a discharge summary or procedure note, dictation is faster and cheaper. The single fastest change for most physicians is writing less, since history and exam no longer drive outpatient E/M code selection [8].

References

FAQs

Related articles