Ambient AI in every department, with notes signed before clinicians leave

The randomised evidence

The first trial randomised 238 outpatient physicians in 14 specialties at one US academic system to one of two ambient scribes or usual care, over two months from November 2024.¹ One scribe reduced time in notes by 9.5% against control, and the other showed no significant change. Physicians in both arms reported clinically significant inaccuracies occasionally, and one mild adverse event was recorded.¹

The second was a 24-week stepped-wedge trial of 66 practitioners across ambulatory clinics in two US states.² Time on notes fell by 0.36 hours a day, diagnostic billing codes improved, and documentation quality scored between 3.97 and 4.99 out of 5 across domains, with no drift in software performance.²

Adoption is the number that moves the result

Both trials were analysed by intention to treat, so their effects include clinicians who rarely used the tool. In the first, the two scribes were used in 33.5% of 24,696 visits and 29.5% of 23,653 visits.¹ In the second, 27,092 of 71,487 notes, or 38%, were drafted with ambient AI.²

Report use by clinician and site every week of a pilot, and find out why low users stopped.

Run a controlled evaluation of your own

Randomise by clinician or use a stepped-wedge design, so every clinician eventually gets the tool and the comparison stays fair.

Fix the primary outcome before launch, usually time in notes from audit logs. Add letter turnaround, coding accuracy checked by coders, clinician-reported inaccuracies and patient experience as secondary measures. Pilot success criteria sets out thresholds that work.

Governance before the second department

WHO guidance on large multi-modal models sets out over 40 recommendations for governments, technology companies and providers, and warns of automation bias, where errors are overlooked or difficult choices are improperly delegated to a model.³ The NIST AI Risk Management Framework, released in January 2023 for voluntary use, gives a common structure for managing those risks across design and use.⁴

Put in place a named clinical safety lead, a written review policy that keeps the clinician as signatory, a consent approach for patients, a monthly audit of signed notes and drift monitoring on every release. The event log shows who drafted, edited and signed each document, and exports on request.⁶

Enterprise procurement that scales

The contract covers security review against your own standards, integration with your record systems, hosting and sub-processors by country, pricing at volume across sites, and exit terms that return every record in a usable format. Name a supplier success lead who owns adoption with you.

wai across the hospital group

wai drafts the note during the consultation, then letters and coding, for the clinician to review and sign, and the same record carries the front desk work forward the moment the note is signed. wai hosts each clinic's records in its own country, under that country's rules. We set this up with your clinic.⁶

wai Enterprise offers volume pricing, security review, custom integrations and a dedicated success manager.⁵ Every message, draft, approval and signature is logged and exportable, and every sub-processor is named with thirty days notice before changes.⁶

Where to start

Pick two departments with different documentation styles and a named clinical lead in each. Agree the primary outcome, the adoption target and the safety measures with your governance group before any recording starts.

Randomise clinicians across a 12 to 24 week window, report use and outcomes weekly, and run the security review in parallel. AI for hospitals covers what to deploy after the scribe.

Frequently asked questions