In June 2026, Canada Health Infoway published the six-month results of its national AI Scribe Program, which enrolled 12,019 primary care clinicians between June 2025 and February 2026. Clinicians saved about 1.1 hours a week during working hours and nearly another hour after hours, and the share of notes finished on the day of the visit rose from 55.9 percent to nearly 75 percent.
Those gains are real, but the note is only one of three pieces of physician paperwork attached to every appointment. Before the visit there is the chart to review, and after it there is the claim to code and submit. Clinics that buy tools for these jobs tend to buy them separately, if they buy them at all, and the research on each suggests the note accounts for only part of the time.
What AI Scribes Have Delivered So Far
The Infoway results are the largest Canadian evidence so far. By February 2026, more than 7,400 clinicians were using an AI scribe at least ten times a month, and monthly AI-assisted encounters had grown from more than 244,000 to more than 1.22 million. Delayed documentation fell from 9.8 percent of notes to under 3 percent, and nearly 70 percent of responding clinicians reported a lighter administrative load. The figures come from the program's own evaluation, and the satisfaction measures reflect the clinicians who answered its surveys.
An earlier provincial evaluation pointed the same way. In OntarioMD's three-month study, published in September 2024, more than 150 family physicians and nurse practitioners reported spending 70 to 90 percent less time on paperwork and saving three to four hours a week, and 83 percent said they would keep using a scribe, all of it self-reported.
Controlled evidence shows smaller gains. In a randomized trial published in NEJM AI in December 2025, 238 outpatient physicians across 14 specialties were assigned to one of two commercial scribes or to usual care. One product reduced time spent in notes by 9.5 percent compared with the control group; the other produced no significant change. Both showed signs of lower burnout that the authors say need confirming in larger trials. Taken together, these studies suggest that scribes help, by amounts that depend heavily on the product and on how it is used, a point we examine in detail in our review of the clinical evidence on AI scribes.
Pre-Visit Chart Review: The Work Before the Visit Starts
A physician who walks into an exam room has usually already spent several minutes in the chart, the review many physicians call pre-charting. In an analysis of electronic record activity by about 155,000 American physicians, published in the Annals of Internal Medicine in 2020 (as reported by Healio), J. Marc Overhage and David McCallie found that physicians spent an average of 16 minutes and 14 seconds in the record per patient encounter. Chart review took the largest share, 33 percent, or five minutes and 22 seconds, ahead of documentation at 24 percent and ordering at 17 percent.
That share has been growing faster than the note itself. Brian Arndt, Christine Sinsky and colleagues followed 141 academic primary care physicians and reported in the Annals of Family Medicine in 2024 that, between 2019 and 2023, their time in the record per eight hours of scheduled clinic rose by 28.4 minutes. Chart review time grew 13.0 percent and inbox time 24.4 percent, while time spent writing notes grew only 2.3 percent. Chart review has been growing more than five times as fast as the note, and a scribe does not address it.
Language models can now draft a summary of the chart before a visit, and a study of a similar task suggests they introduce more errors than physicians do. In a 2025 study in JAMA Internal Medicine, physicians reviewing hospital discharge summaries found an average of 2.91 errors in each narrative drafted by a large language model, against 1.82 in those written by physicians. Reviewers rated overall quality about the same, and potential harm was low in both groups, though higher in the narratives the model wrote. For a summary read in the minute before a visit, two safeguards matter most: every statement should point back to where it came from in the chart, and the summary should report what the chart says rather than interpret it.
Physician Billing After the Note: Where Revenue Slips
Once the note is done, the visit still has to be paid for, and that step loses money in ways a clinic rarely sees in full. The Ontario Medical Association reported in March 2026 that OHIP flags more than one million of the roughly 200 million claims it processes each year, and that about 58,000 challenged claims a year can take months or even years to be paid.
Billing is also a large share of the paperwork itself. In an October 2025 survey of nearly 500 Canadian physicians, run by the billing platform Dr. Bill, respondents reported spending about 7.5 hours a week on non-clinical work and named billing as the single most time-consuming non-clinical task. Seventy-two percent said administrative duties contribute significantly to burnout.
General-purpose models have not shown they can choose codes reliably on their own. In an April 2024 study in NEJM AI, researchers at Mount Sinai found that GPT-4, the best of the models they tested, reproduced the exact code from a description only 45.9 percent of the time for ICD-9-CM diagnoses and 49.8 percent of the time for CPT procedures. A 2026 claims analysis by Trilliant Health found that, at six American health systems, the share of visits billed at the highest levels rose after they adopted AI scribes. The authors say the causes are not clearly understood and consider fraud an unlikely explanation, but the finding shows that a change in how notes are written changes the claims that follow them. The safer design assigns codes from what the note actually documents and shows the physician why each was chosen before anything is submitted.
AI Scribe Consent and Physician Review in Canada
Canadian medical-legal and privacy authorities agree on two requirements. The Canadian Medical Protective Association's guidance on AI scribes, revised in December 2025, advises physicians to obtain the patient's consent before recording a visit, to explain what the recording is for, to document that discussion in the chart and to review every note the scribe drafts. Ontario's Information and Privacy Commissioner, in its January 2026 guidance on AI scribes, says consent would generally be required and that a person must review AI-generated content before it is entered into the record. A claim goes out under the physician's billing number, so the same review should apply to it. Alberta's privacy commissioner has published privacy impact assessment guidance for custodians who adopt one.
We cover the provincial rules in more depth in our articles on Ontario's AI guidelines, Alberta's AI guidance and British Columbia's AI scribe guidance.
One Chart, Three Jobs: How the Brief, the Note and the Claim Connect
The patient's chart is the record that ties the visit together, and each of these tools is only as good as its connection to it. The front desk books the appointment and updates the patient's details in the chart. A pre-visit summary reads that same chart, including the reason for the visit, recent results and the letters from other physicians, so the physician starts from the current record rather than the last time they looked. The scribe's note goes back into the chart once the physician approves it, and the claim is built from the note rather than from a second pass through the visit.
On separate products, a summary can miss the referral letter filed yesterday, the note is rekeyed into the record, and the billing waits for the end of the week. When each tool reads from and writes to the same chart, what one step records is in the chart for the next step that needs it. The front-desk side of this is in our guide to confirmations, pre-visit questionnaires and follow-ups, and the general case in The Point Solution Trap.
How JOUD Supports the Physician From Brief to Claim
JOUD covers the physician's side of the visit as well as the front desk, with one sign-in and one clinic setup for both. Nothing reaches the chart or Alberta Health until the physician approves it. The scribe works with TELUS CHR, and billing covers Alberta claims.
- Before the visit, it prepares a one-screen brief of the patient's chart: allergies, alerts, the reason for today's visit and what has changed since the last one, active problems, medication changes, recent results with their trend, overdue preventive care, and referrals and letters. Every line names where in the chart it came from, anything it cannot trace to the chart is left out, and the brief reports facts without adding diagnoses or suggestions.
- During the visit, it records the conversation, separates the physician's voice from the patient's, and drafts the note in the sections the clinic's TELUS CHR chart uses, or in SOAP format if the physician prefers. Every section can be edited before approval.
- After the visit, the physician approves the note and places it into the patient's TELUS CHR chart in one click. A visit can be shared with colleagues at the clinic, who can view it but not change it.
- For billing, it suggests Alberta fee and diagnostic codes based on what the note documents and shows the evidence behind each one, including the alternatives it considered. The physician confirms the time spent, and the billing updates to match. The patient's health number and birth date are checked with the billing service before any claim is created.
- Once approved, claims go to Alberta Health through the clinic's billing service, and each one is tracked through acceptance, holds, refusal or payment, with Alberta Health's explanation codes shown when a claim is refused or held. The clinic sees what it expects to be paid, what is outstanding and what has arrived.
Elevation Labs builds clinic operations AI for primary care and specialist clinics across Canada and the United States. If your physicians are finishing notes at night and entering billing at the end of the week, book a demo and we will show you a visit from brief to claim.
Frequently Asked Questions
Do I need patient consent to use an AI scribe in Canada?
In most cases, yes. Ontario's privacy commissioner says consent would generally be required, the CMPA advises obtaining it before any recording, and Alberta's privacy commissioner has published privacy impact assessment guidance for clinics that adopt a scribe. Check the guidance from your own province's privacy commissioner and college as well.
What does the CMPA say about AI scribes?
The Canadian Medical Protective Association's guidance, revised in December 2025, also asks physicians to explain the privacy and accuracy risks and to keep any consent form with the chart. It advises them to obtain the patient's consent before recording, explain what the recording is for, document that discussion in the chart and review every note the scribe drafts.
Can an AI scribe do my billing?
It can suggest codes from what the note documents, which spares the physician from coding the visit from memory later. Because general-purpose AI models have struggled to choose exact codes, the physician should see why each code was suggested and approve the claim before it is submitted.
What is a pre-visit chart summary?
It is a short brief of the patient's chart prepared before the visit, the work many physicians call pre-charting, covering items such as allergies, active problems, medication changes and recent results. A good one shows where each statement came from in the chart and does not add interpretations of its own.
How much time do physicians spend reviewing the chart before a visit?
In an analysis of about 155,000 American physicians, chart review took five minutes and 22 seconds of the 16 minutes and 14 seconds spent in the record per encounter, the largest share of that time.
Does an AI scribe reduce chart review time?
Not directly. A scribe drafts the note from the conversation, while chart review happens before the visit. In a study of 141 academic primary care physicians, chart review time grew 13.0 percent between 2019 and 2023 while note writing grew 2.3 percent, which is why a pre-visit summary is a separate tool.