Webinar: Unlocking the Potential of AI Scribes

Zayed offered a 5-level hierarchy for understanding how AI automation can be used to automate healthcare tasks.

The market’s current focus lies on lower-level automation due to the complexity and ethical considerations involved in higher levels, with level 3 applications (ones that provide clinical decision support) being an underexplored yet promising category for impactful care improvements.

“I think there’s an aspiration pretty universally that AI scribes are going to move from level 2 (administrative and backend automation) up towards level 3 (clinical decision making) and starting to do more and more, but it’s really not clear at what pace that aspiration is going to play out.” (Zayed)

Read more on Zayed’s framework here

Our panelists categorized the AI scribe landscape into three major groups: emerging startups, established startups (e.g., Ambience, Abridge, DeepScribe), and larger entities like Nuance DAX (from Microsoft).

Each category presents distinct offerings and capabilities. However, our panelists were wary of newer solutions that simply apply a wrapper to a chatGPT: “Go with a vendor whose existence preceded chat-GPT,” Zayed recommended, arguing that many AI scribes born in the chat-GPT boom don’t build with the clinician-patient interaction in mind.

Our panelists noted the importance of constructing a financial case based on clinician productivity gains (i.e. ability to see additional patients) in building organizational buy-in for a potential AI scribe investment.

However, they also cautioned not to get carried away by the expected revenue impact from these solutions, emphasizing instead what they saw as the true primary benefit to be measured: clinician burnout mitigation and overall wellness improvement.

As Kendell put it, “When you need to sell it to a CFO, it’s a pretty easy financial calculation to see how an AI scribe enables you to see one additional patient a week, but from the clinical perspective, that is not what we’re trying to achieve….that is not the goal of this.”

Prioritizing comprehensive solution evaluations is key when considering AI scribe implementations as vendor assertions may not reflect real-world effectiveness. Panelists highlighted that relying solely on recorded demos is insufficient and that reviews, live demos, and trials provide a more complete understanding of product functionality.

“It’s one thing to watch a demo, but if you are brave enough to say ‘we will let you try this out in the wild,’ that’s way better because things are different in a conference room than they are in an exam room,” Matt recalled.

Differentiating factors among AI scribe solutions primarily revolved around accuracy, cost, transparency, and integration with existing Electronic Health Record (EHR) systems.

However, certain panelists highlighted the absence of clear frontrunners and noted that some of the key distinguishing factors had not yet surfaced. Matt observed, “I haven’t seen any vendors that have blown the others out of the water by a huge margin.”

There’s varying perspectives on standardizing note-taking styles across departments for ease of maintenance and training versus allowing individual clinicians to customize notes based on personal preferences.

Kendell highlighted her aim to standardize documentation processes across her practice, preferring a standardized length for documentation, whereas Matt was more interested in solutions that could be tailored to individual clinician preferences for longer, detailed notes or bullet points.

The AI scribe landscape is rapidly evolving, with certain vendors adding lots of capabilities and transforming from mere point solutions into comprehensive platforms. What held true a year ago may not align with today’s reality, so thorough and consistent vendor evaluations are imperative in such a dynamic landscape.