Ambient Scribing - Trends Impacts Clinical Documentation Workflows

HIPAA compliant ambient scribing has emerged as a valuable tool in helping healthcare organizations address the crisis of physician well-being and burnout. As such, it sits at the intersection of two of the most critical trends impacting healthcare in recent years — the epidemic of physician burnout, and also the AI developments that portend key change ahead, according to healthcare clinical and IT leaders.

Each day, on average, providers spend almost two hours on clinical notes according to a study noted in the Annals of Internal Medicine — often during evening hours, completing notes instead of investing time with family and their own wellness. The study notes only 27% of the day is spent on facetime directly with patients, for physicians in outpatient settings. In this environment, and coming off the back of the COVID epidemic, 53% of physicians in a Medscape 2023 survey reported feeling burned out. Supporting physicians has become a topline consideration for organizations, as they expand access to care amidst resourcing and reorganization challenges. Last year, the NIH noted that physician turnover costs have ranged from $88,000 to $1,000,000 per physician.

Enter ambient scribing. It gives time back to the providers, by using healthcare-tuned AI to generate drafts of clinical notes for clinicians to review, edit, and approve.

In this post, we’ll look at functionality driving adoption of ambient scribing. We’ll also discuss key considerations for telehealth providers, organizations and leaders as they evaluate these applications. Ambient scribing brings to the fore questions of security, AI strategy, and technology pathways in a rapidly evolving landscape.

##A Simple Experience — with Clinician as Pilot

The ambient scribe workflow itself is straightforward — the very purpose is for the technology to operate behind the scenes. The provider can focus on talking to the patient, instead of taking notes.

As the provider talks to the patient, the device microphone captures the conversation. Often this is thought of as a “recording” — though technically, a recording is not required. (State-of-the-art solutions instead use real-time speech-to-text (STT) transcription.) It’s key that the solution be able to work in a noisy environment and support diverse provider needs and diverse patient populations, as well.

For example, leading solutions typically provide functionality like:

  • Multiple languages – support diverse patient populations with ambient support for an array of languages
  • Multiple templates – the provider can select the structure of the note, related to their specialty
  • Editing panel – it’s key for the provider to be in the driver’s seat, and be able to review and edit the AI-generated draft.
  • EHR integration – for the most seamless experience, the AI-generated draft is sent straight into the EHR, for physician approval there.
SAAS example of transcription

Leading solutions generate documentation in real-time. Once the conversation ends, the clinician can see the documentation the AI has generated. The ambient scribe supplements and supports the clinicians, setting them up for success.

Click here to watch a short video.

##The role of telehealth

Before diving into AI considerations, it’s worth first noting the importance of the telehealth perspective with regards to scribing in general. As organizations work to expand access, telehealth has become a key node of care. (Over 80% of Americans have seen a provider virtually.)

The very benefit of the ambient scribe is its seamless input into the system of record. In considering their needs and a path forward, an organization should consider whether the ambient scribe supports both in person and virtual encounters. If the same ambient listening can be deployed in a range of settings, this offers various benefits —from time savings wherever care is provided; to operational efficiency (like training, support, procurement) in working with a consistent vendor; and to enabling a consistent approach to documentation, within the patient chart.


Prior to generative AI, healthcare for several years has explored how natural language processing (NLP) could help ease this documentation burden — for example, early adopters used dictation software in clinical practice.

The November 2022 launch of ChatGPT — and DALL-E and Whisper — reframed this technology discourse around generative AI, and the large language models (LLMs) and foundation models undergirding it.

The last couple years has seen a startling pace of technology development. One key trend has been the rollout of purpose-built LLMs and language models trained on industry-specific data sets.

Healthcare has been at the forefront of this development. Enterprise and academic organizations have moved to leverage data insights that align with their organization strategies and roadmap.

  • Becker Editor-In-Chief Laura Dyrda recently talked with Scott Becker on how systems are assessing how to leverage their current partnerships as they review generative AI roadmaps. For example, Mass General Brigham and GE HealthCare are adding foundational model research into their strategic 10-year AI collaboration plan.
  • Reflecting on academic systems, University of Rochester Chief Digital Health Officer Michael Hasselberg noted their full integration with the university, and the private access it affords them to foundation models and insights and expertise from the data science faculty.
  • In industry, Veradigm earlier this year acquired ScienceIO, the AI platform and foundation model provider for healthcare. According to Interim CEO Dr. Yin Ho, the acquisition enables Veradigm to create its own proprietary LLMs on Veradigm’s data set, spanning over 400,000 providers and 200 million patients.

These developments point to two key takeaways. First, for healthcare organizations, it is a standard requirement that the ambient scribing solution leverage healthcare-tuned AI. The ambient scribing space is competitive and well-provisioned with many vendors. It is key that any ambient listening solution — or any AI tooling — use state-of-the-art language, healthcare-tuned models.

Second, beyond leveraging a product’s default healthcare model, organizations should consider the modularity of the AI model the solution deploys. Specifically, enterprise organizations that own large data sets, and are evaluating foundational model strategies, should work with vendors whose architecture lets them swap in models. The ambient scribing solution should seamlessly enable rollout out of proprietary workflows and algorithms.

The solution, optimally, is architected for flexibility, to support the AI strategy of the healthcare organization — instead of the organization being locked into the product’s AI strategy. This underscores the difference between vendors that only provide ambient scribing as a product, versus vendors that support both product and platform.

##Technology pathway

The ambient scribing space is competitive and well-provisioned with a variety of vendors. The majority of these vendors provide essentially off-the-shelf experiences. A healthcare organization buys the product and deploys it with standard functionality. This off-the-shelf approach can provide the essential functionality, to help give back time to clinicians.

This “buy” approach is one end of the typical procurement spectrum. At the other end is a “build” approach. This could be a fit for strategic entities like EHRs. This approach likely is not a fit for healthcare organizations that deliver services directly into a patient population.

There is a third, more flexible option that either build or buy. A full stack vendor provides both off-the-shelf solutions, as well as the platform flexibility to support the customizations that matter to clinicians. Customizable templates are a good example of the importance of flexibility. Most products offer multiple templates — a provider can pick a template suited to their specialty, like SOAP or DPA notes, or documentation for a specialty like cardiology.

Customizable templates give organizations even more alignment. The organization can work with the vendor so that the template reflects how the clinicians and systems work. This full stack experience puts a product in the hands of a customer immediately, while also positioning them better for long term success, with support for customization and API-driven workflows.

##HIPAA compliance

Finally, it is worth, as always, stressing compliance adherence. While HIPAA compliance should be a default requirement, it is important to note that AI will introduce heightened scrutiny. It is imperative that organizations understand where their data is being sent for training.

Beyond checking the HIPAA box, organizations should ask where engineers are located. It is also imperative to understand corporate governance, and what foreign governments have access to any data. (For example, any corporate entity with headquarters in China is required to turn over data to the Chinese government — this requirement is true in nearly most countries, but it is useful to clarify the legal process to which your data is exposed.)


Ambient scribing has seen keen interest and genuine enthusiasm from providers and systems. In making an impact on what matters most —clinicians and quality care for patients —its immediate, tangible benefits are exciting. It is also an interesting contrast to previous technology deployments, which unintentionally have created more tooling burden. Responsibly deployed and aligned with industry safeguards and standards, AI, as reflected in ambient scribing, is a marked change. Telehealth clinicians and leaders have an exciting opportunity to deliver change in the moment, and by considering overall trends and strategies, deliver long term success, for clinicians and patient care.