Finish clinic with the note and the letter already signed, and choose the scribe that gets you there

Six outputs from every consultation

Every consultation produces a note, a letter to a colleague or the patient, codes for the claim, often a clinical question answered mid-visit, a follow-up for the front desk and an update for the patient's own record. Score all six, plus the audit trail.

The note, judged on what it misses

In a three-arm randomised trial of 238 outpatient physicians across 14 specialties at one US academic health system, one scribe reduced time in notes by 9.5% against usual care, while the other showed no significant change.² Physicians using either tool reported clinically significant inaccuracies occasionally.²

Run every tool on the same recorded consultations of your own. Time how long each draft takes to check, and look hardest at omissions and at anything inferred that was never said. Our guide to AI scribe accuracy sets out the review routine.

Letters that are ready to sign

Test each tool on a referral that needs a history and a clear question for the colleague receiving it. A strong draft comes from the same consultation, in your template and addressed correctly. Count the edits before you would sign it, and pass it at three or fewer. The referral letter explains why this document carries the most weight.

Codes and cited evidence

In a stepped-wedge randomised trial of 66 practitioners in two US states, diagnostic billing codes improved with ambient AI and documentation time fell, with no loss of note quality or billing compliance.³ Ask each tool to suggest codes from your test consultations and have your coder mark them against the signed note.

In a 2026 survey of 1,342 US physicians, 39% used AI to summarise research and standards of care, the most common clinical use.¹ Pass an answer only when every claim links to guidance or a paper you can open.

The front desk and the patient's own record

In an AMA survey of 1,000 US physicians, practices completed 39 prior authorisation requests per physician each week, taking physicians and staff about 13 hours.⁴ Check that the follow-up, the booking and the claim start from the signed note without retyping.

Then check what the patient receives. Across OECD countries, 90% have an electronic health portal in place, and 42% report that people can both access and interact with all their data through it.⁵ Pass the tool when the signed letter and results reach the patient's own record, with permissions the patient controls.

The whole consultation on wai

wai Clinicians drafts the note during the consultation, then the letters, suggested codes and result explanations, all held for the clinician to review and sign.⁶ The same record feeds wai Admin, where the team approves drafted follow-up, booking and claims prepared for insurers.⁷

Patients read their letters and results in their own wai Profile. Prices are on the pricing page, and a seven day trial for one clinician, with limited use, is enough to start the scorecard on real consultations.

Where to start

Choose five recent consultations that include a referral, a result discussion and a new patient. Agree the pass mark for each row of the scorecard before any tool sees them.

Run every shortlisted tool on the same five. Time your review of each output, count the edits, and ask your coder and front desk to score their own rows. Total the scores, then take the contract questions to how to choose an AI medical scribe.

Frequently asked questions