Microsoft has launched a brand new voice-activated AI assistant designed to assist docs and healthcare professionals transcribe medical notes and draft paperwork in addition to shortly seek for info from medical sources.
The brand new healthcare AI device known as Dragon Copilot is being supplied as a part of Microsoft Cloud for Healthcare. It harnesses the pure language voice dictation and ambient listening know-how developed by AI voice firm Nuance, which was acquired by Microsoft for $16 billion in 2021.
These capabilities have been additional fine-tuned utilizing generative AI and tailored to include healthcare safeguards, the corporate mentioned in a weblog publish printed on Monday, March 3.
The rise of generative AI instruments for the healthcare trade is pushed by the necessity to streamline administrative duties, enhance effectivity, and fight burnout amongst clinicians. Microsoft has emerged as a significant participant within the fast-growing marketplace for AI note-taking instruments.
“Nobody turns into a clinician to do paperwork, however it’s turning into a much bigger and greater administrative burden, taking time and a focus away from truly treating and supporting sufferers. That’s why we’re introducing Microsoft Dragon Copilot, the trade’s first AI assistant for medical workflow,” Microsoft CEO Satya Nadella mentioned in a publish on X.
Right here’s a better take a look at how Dragon Copilot works, what it might probably do, and the dangers posed by healthcare AI instruments.
What’s Dragon Copilot?
Microsoft Dragon Copilot is constructed on high of present instruments similar to Dragon Medical One (DMO) and DAX rolled out by speech recognition firm Nuance Communications.
Story continues under this advert
In response to Microsoft, DMO’s speech capabilities has helped clinicians transcribe “billions of affected person information” whereas DAX’s ambient AI know-how has “assisted over 3 million ambient affected person conversations throughout 600 healthcare organisations prior to now month alone.”
The underlying structure of Dragon Copilot permits organisations to ship enhanced experiences and outcomes throughout care settings for suppliers and sufferers alike, Microsoft mentioned.
“With the launch of our new Dragon Copilot, we’re introducing the primary unified voice AI expertise to the market, drawing on our trusted, decades-long experience that has persistently enhanced supplier wellness and improved medical and monetary outcomes for supplier organizations and the sufferers they serve,” Joe Petro, the company vp of Microsoft Well being and Life Sciences Options and Platforms, mentioned in a press release.
How one can use Dragon Copilot?
Dragon Copilot can be utilized to draft memos and notes in a personalised model and format, as per Microsoft. Apart from voice-to-text transcription, the Dragon Copilot person interface additionally permits customers to submit prompts or use templates for creating AI-generated notes.
Story continues under this advert
Aside from documentation work, the AI assistant permits clinicians to seek for general-purpose medical info from trusted sources. It may also be used to “automate key duties similar to conversational orders, notice and medical proof summaries, referral letters, and after-visit summaries, in a single centralised workspace,” Microsoft mentioned.
To substantiate its claims of Dragon Copilot’s utility in healthcare areas, Microsoft supplied findings from its survey which revealed that the AI assistant helped clinicians save 5 minutes in each affected person interplay. “Round 70% of clinicians reporting diminished emotions of burnout and fatigue, 62% of clinicians said they’re much less more likely to go away their group, whereas 93% of sufferers report a greater total expertise,” Microsoft mentioned.
Dragon Copilot will reportedly be accessible via a cellular app, browser or desktop, and it integrates immediately with a number of completely different digital well being information. The AI assistant will grow to be out there to be used within the US and Canada in Could this yr. Quickly after, will probably be launched within the UK, Germany, France, the Netherlands, and different key markets.
Nonetheless, Microsoft didn’t present particulars on the price of accessing Dragon Copilot.
Story continues under this advert
Rise of AI instruments in healthcare
Startups and tech firms are racing to introduce generative AI instruments and {hardware} for the healthcare trade. On Monday, Google Cloud printed a weblog publish that laid out how healthcare suppliers are utilizing the tech big’s choices in medical settings. Gamers similar to Basalt Well being are utilizing Vertex AI, Gemini, and different Google Cloud applied sciences to create AI brokers that help medical assistants in getting ready affected person charts and performing administrative duties in addition to flagging potential well being dangers.
Startups similar to Abridge and Suki have raised over $460 million and practically $170 million, respectively, for growing comparable AI scribing instruments.
Nonetheless, generative AI instruments in healthcare carry the identical danger of huge language fashions (LLMs) hallucinating or making issues up. This danger turns into much more vital in a medical setting the place affected person lives are probably on the road.
Hidden dangers of healthcare AI instruments
The Related Press discovered that OpenAI’s Whisper device for transcribing docs’ consultations with sufferers, was vulnerable to hallucinations. Whisper made up chunks of textual content that included racial commentary and violent rhetoric in addition to imagined medical remedies, the report claimed citing interviews with a number of software program engineers, builders, and tutorial researchers.
Story continues under this advert
Whereas acknowledging the potential purposes of AI units in healthcare, the US Meals and Drug Administration (FDA) recognized hallucinations in AI fashions as a problem.
“For instance, for a GenAI-enabled product that could be meant to summarize a affected person’s interplay with a well being care skilled, the potential for that product hallucinating
can current the distinction between summarising a healthcare skilled’s dialogue with a affected person and offering a brand new prognosis that was not raised in the course of the interplay,” the FDA mentioned in a report final yr titled ‘Whole Product Lifecycle Issues for Generative AI- Enabled Gadgets’.
It additionally identified that foundational AI fashions “could also be vulnerable to bias that could be
particularly tough for particular person product builders to establish or mitigate for his or her ensuing GenAI-enabled merchandise.”
On Dragon Copilot, Microsoft has mentioned that “healthcare-specific medical, chat and compliance safeguards for correct and secure AI outputs” have been integrated into the AI assistant. The corporate added that Dragon Copilot has been developed according to Microsoft’s accountable AI rules.
Story continues under this advert
It has not elaborated on how the healthcare AI device addresses dangers of hallucination and efficiency bias.
