Open-Source AI Scribe 'Berta' Cuts Clinical Documentation Costs by Up to 95% in Canadian Provincial Deployment
A preprint study published on arXiv details Berta, an open-source AI clinical documentation tool deployed across Alberta Health Services, achieving operating costs under $30 per physician monthly — up to 95% less than commercial alternatives — while retaining full data sovereignty within the provincial health system.
A new open-source AI clinical documentation tool called Berta has demonstrated dramatic cost savings and rapid adoption across Alberta's provincial health system, according to a preprint study published on arXiv (https://arxiv.org/abs/2603.23513).
Developed as a direct response to the high cost and closed nature of commercial AI scribes — which typically charge between $99 and $600 per physician per month — Berta was deployed within Alberta Health Services (AHS) and integrated with the organization's existing Snowflake AI Data Cloud infrastructure. The system combines automatic speech recognition with large language models while keeping all clinical data within AHS's secure environment, a design choice the researchers describe as essential for data sovereignty and institutional governance.
Over eight months, from November 2024 to July 2025, 198 emergency physicians across 105 urban and rural AHS facilities used the platform, generating 22,148 clinical sessions and more than 2,800 hours of audio. Monthly usage grew from 680 sessions to 5,530 — an increase of more than 700 percent — over the study period. Operating costs averaged less than $30 per physician per month, representing a 70 to 95 percent reduction compared to commercial alternatives, according to the paper.
AHS has since approved expansion of the program to 850 physicians, which the researchers describe as the first provincial-scale deployment of an AI scribe integrated with existing health system infrastructure.
The paper highlights a structural criticism of the commercial AI scribe market: that dominant vendors operate as opaque systems that do not return data to institutional infrastructure, limiting health organizations' ability to oversee quality, customize workflows, or maintain control over sensitive patient information. Berta, by contrast, is being released publicly as open-source software so other health systems can adapt it to their own environments.
This deployment arrives amid a broader wave of AI adoption in healthcare administration, where ambient documentation tools have attracted significant venture investment and clinical interest. The Berta study adds an important counterpoint to that trend — suggesting that institutional control, transparency, and cost-effectiveness are achievable without relying on proprietary commercial platforms. Whether other health systems will have the technical infrastructure and institutional will to replicate Alberta's approach remains an open question.
PRAXIS: Regular patients could soon notice doctors spending more time listening and less time typing, while public health systems stretch their budgets further without handing sensitive records to big tech firms. This points to a future where useful AI quietly becomes a shared community tool instead of an expensive subscription.
Sources (1)
- [1]Berta: an open-source, modular tool for AI-enabled clinical documentation(https://arxiv.org/abs/2603.23513)