Key points
- CMS lets a billing specialist or alternate source review documentation and help choose codes, and the clinician still makes sure the record supports the level billed.1
- Since 1 January 2023, most E/M visit levels rest on medical decision making or total time, and CMS says the volume of documentation must not decide the level.1
- CMS attributes 77.17% of fiscal year 2025 Medicaid improper payments to insufficient documentation, generally not fraud.3
- wai drafts suggested coding from the consultation for the clinician to review and sign, and every draft and signature is logged.67
Summary
The clinician who bills a code is accountable for it, whoever suggested it. Accept an AI-suggested code when the signed note shows the decision making or time behind it, and every claim carries its own proof.
Every code you bill is one you chose, with the signed note that supports it on the same page. The note is drafted during the consultation and the code is suggested from it, so you approve both in one sitting. The billing rules stay the same, and they are easy to meet when the evidence for each code sits beside it.
See it in your clinicThirty minutes on your own patients and doctors.
Request a demoSee a consultation become a signed note with its suggested codes beside it, and the log of who approved what.
| Check | What the note should show | Basis |
|---|---|---|
| Level basis | Whether the level rests on medical decision making or on total time | CMS E/M guidance, from 1 January 2023 |
| Decision making | Problems addressed, data reviewed and risk of the management chosen | AMA CPT E/M guidelines |
| Time | Total time on the date of service, with the service provided for the full time | CMS E/M guidance |
| Medical necessity | The sign, symptom or complaint that makes the service reasonable and necessary | CMS E/M guidance |
| Diagnosis codes | Diagnoses in the assessment that match the codes on the claim | CMS documentation principles |
| Length | Nothing, on its own, since volume of documentation must not set the level | CMS E/M guidance |
| Signature | Your review and signature, after any change to the code | Clinician accountability |
The code belongs to the clinician who signs
CMS's evaluation and management guidance, updated in May 2026, says a billing specialist or alternate source may review documentation before the claim is sent and help choose codes.1 It still requires the clinician to make sure the claim correctly shows the services, that the record supports the level reported, and that the volume of documentation does not decide the level.1
An AI suggestion fits that model as a fast reviewer. Read the suggestion against the note, change it if needed, and sign. The same principle governs the note itself, covered in who signs an AI clinical note.
What the note has to show
Since 1 January 2023, most E/M visit families are levelled on medical decision making or on total time on the date of service.1 History and examination must be medically appropriate and no longer set the level.12 The AMA frames the change as documentation centred on how physicians think, with time defined as minimum time.2
CMS expects each note to give the reason for the encounter with relevant history and findings, then the assessment and the plan of care.1 The record should reflect the diagnosis and treatment codes on the claim, billing a higher level when a lower one fits is not medically necessary, and when time is used the service must be provided for the full time.1
Length is not level
CMS says the volume of documentation must not decide the level billed.1 A thorough ambient note for a straightforward visit supports a straightforward code.
When the suggested level sits above the decision making on the page, lower the code or complete the record with what genuinely happened. Coding below what the note supports leaves earned revenue unbilled, as set out in our guide to AI coding and undercoding.
Early evidence on scribes and billing
A JAMA Network Open study at UCSF Health followed 1,565 physicians and 1.2 million ambulatory encounters from 2023 to 2025. The 698 who adopted an AI scribe billed 1.81 more RVUs a week than non-adopters and saw 0.80 more patients a week, with no difference in the proportion of claims denied.4 The authors call for research on whether the extra RVUs reflect more services or more accurate coding.4
A randomised trial of 66 practitioners in Wisconsin found that billing diagnostic codes, reviewed by professional coders, improved with ambient AI, while documentation time fell without compromising billing compliance or note quality.5
Ready for audit by design
Medicare's fee-for-service improper payment rate was 6.55% in fiscal year 2025, or $28.83 billion, and CMS says most improper payments across its programs arise where documentation is missing or insufficient.3
A Medicare contractor may ask the record to confirm the site of service, the medical necessity of the service and that it was reported correctly.1 Checking each suggested code against the note at signing answers that request in advance.
Each month, pull ten encounters at random, set the suggested code, the billed code and the note side by side, and record any change with its reason. Outside the US the code sets differ, and the same audit applies.
How wai handles coding
wai drafts the visit note during the consultation and prepares letters and suggested coding from it for the clinician to check and sign.6 Only the clinician signs a clinical document, and nothing clinical goes out without that signature.67 Every draft, approval and signature is written to an event log the clinic can export, so each code's path from suggestion to signature is on record.7
Where to start
Decide, per visit type, whether you level on decision making or on time, and make sure the note template captures that basis every time.
For the first month with AI coding, check every suggestion against the table below before signing. Then run the ten-encounter audit monthly and track the share of suggestions you change and your denial rate.
Bring three de-identified visit types from your own list. We show the note, the suggested codes and the record of your sign-off.
Frequently asked questions
If an AI tool suggests the wrong code, who is responsible?
The clinician who signs and bills it. CMS guidance allows a reviewer to help choose codes but requires the clinician to make sure the documentation supports the level of service billed.1
Does a longer AI note justify a higher E/M level?
No. CMS says the volume of documentation must not be used to decide the level of service. The level rests on medical decision making or total time.1
Do AI scribes increase claim denials?
In a study of 1,565 physicians at UCSF Health, AI scribe adopters billed more RVUs with no difference in the share of claims with a denial.4 Track your own denial rate from the first month.
Can a clinician accept AI codes without reading them?
Read each suggestion against the note before signing. A monthly audit shows how often suggestions change, and the reading continues after the numbers settle.
What should a monthly coding self-audit include?
Ten random encounters, each with its suggested code, billed code and note side by side, plus any change and its reason. Add the month's denial rate.
Sources
- 1Centers for Medicare and Medicaid Services, 2026Evaluation and Management Services, MLN Booklet MLN006764Read 14 October 2026Read source
- 2American Medical Association, 2023CPT evaluation and management (E/M) code and guideline changesRead 14 October 2026Read source
- 3Centers for Medicare and Medicaid Services, 2025Fiscal Year 2025 Improper Payments Fact SheetRead 14 October 2026Read source
- 4JAMA Network Open, Holmgren and colleagues, 2026Ambient Artificial Intelligence Scribes and Physician Financial ProductivityRead 14 October 2026Read source
- 5NEJM AI, Afshar and colleagues, 2025A Pragmatic Randomized Controlled Trial of Ambient Artificial Intelligence to Improve Health Practitioner Well-BeingRead 14 October 2026Read source
- 6wai, 2026AI medical scribe for cliniciansRead 14 October 2026Read source
- 7wai, 2026Security at waiRead 14 October 2026Read source
Every code you bill, chosen by you and proven by your note.
Fifteen minutes on video with a clinician who uses wai daily, from consultation to signed note and suggested codes. We reply within one working day.
Partner with us
See it on your own patients, then build it with us
Thirty minutes on your own patients, doctors and rules, in your market. We reply within one working day, and the clinics that partner with us early shape what we build next.
See what wai does between the visits