AI scribe predictions
Hey everyone,
I think a new AI scribe unicorn is inevitable, there is a lot of competition in this space and new startups continue to emerge. Before I share my predictions, let's break down the parts of an AI scribe.
There are two components to an AI scribe:
Step One
Accurately transcribing/ translating the clinical visit, this is largely a commodity with several off the shelf Large Language Models. A few examples being ChatGPT by OpenAI, Gemini by Google and LLama by Meta.
Most startups have no IP here and are using one of the above existing solutions.
Step Two
Converting this verbatim transcription into a usable record. Generally this is in the format of a SOAP (subjective, objective, assessment and plan) note. In addition to the transcription we often pull health data (imaging, labwork, vital signs, previous consults etc) to complete our note.
There is room for creativity in the format itself as some of us are moving away from the SOAP format.
Differentiation
Who ever focuses on step two and separates themselves from competitors will win the AI scribe war. While there are early market leaders like Abridge, DeepScribe, Ambience and Nabla. Newer competitors such as Heidi, Scribbery and Freed are quickly capturing market share.
Accuracy is less important here than usability.
The winner will understand this and leverage already existing EMR data to provide a note which not only captures the clinical encounter but integrates the patients history, their medical tests and previous consults to create a succint, clear and comprehensive note.
I am still waiting to invest in founders who are working an AI scribe which does the above.
Thank you for reading,
Rishad