Skip to content
Fast Company Executive Board

The Fast Company Executive Board is a private, fee-based network of influential leaders, experts, executives, and entrepreneurs who share their insights with our audience.

Healthcare AI will succeed as a scalpel, not a sledgehammer

For AI to really deliver value at the patient, population, and clinician level, it requires a thoughtful, precise approach that is deeply integrated into the tech stack with a scalpel to revolutionize the experience of care from within. 

Healthcare AI will succeed as a scalpel, not a sledgehammer

[Parichat / Adobe Stock]

Profile photo of Kyna Fong

CEO and co-founder · Elation Health

After a year of working on implementing AI into our electronic health record (EHR) platform, I find myself in a position of some optimism and some pessimism. AI continues to evolve at breakneck speed, making it almost impossible to claim true expertise. However, we have learned a few valuable lessons from working closely with our community of primary care clinicians and I hope sharing these learnings can carry forward to the rest of the health technology landscape in the year ahead.  

On the side of optimism: AI scribe technologies, tools that listen to the physician-patient conversation and produce detailed notes, are already making a big impact on documentation work for medical professionals. In a recent survey of Elation customers, 66% of our primary care clinicians reported that this kind of AI had “significantly decreased” their administrative burden, and nearly the same high percentage said they expected AI to improve clinical outcomes and practice revenues in the next year. This is a massive step forward, and what we have hoped for with implementing AI in our platform. 

Outside of our community, widespread adoption of AI scribe systems by clinicians using various EHRs this year has been shown to significantly reduce administrative time outside the exam room—combatting clinician burnout—as well as boost patient perception of physician attention and enable consultations in the patient’s own language to be transcribed into English in real time. 

Yet, the implementation of AI into the EHR system is a sticking point. The same survey found that “integration with existing systems” was the leading reason selected by physicians who didn’t feel like AI had reduced their documentation burden yet. We’ve experienced firsthand these challenges of successfully integrating third-party AI into core workflows in a seamless fashion—it is a delicate nuance to get right.

Clinicians in a survey of Kaiser Permanente Medical Group reported similar feelings, saying that “there was still a significant amount of editing that had to be done” and that AI might not be quite ready for prime time. AI scribes are producing fairly high-quality transcripts of conversations, but clinicians are often still having to work to translate lay conversation (tummy ache) into organized professional documentation (abdominal pain), to move it into the structured fields of the EHR, to identify the action items, and to execute those orders accordingly.

With those counterpoints together, what we’ve learned this year is that the real success of healthcare AI appears to be in implementation with a scalpel, not a sledgehammer. The broad hype of AI systems replacing physicians is not where we see promise. The creation of large new AI use cases—such as diagnostic systems—which can stand up a whole company around them is one area of opportunity, but our takeaway is that it’s in the small stuff that AI can truly shine. 

As we look to the future of AI development in the next year, what’s most exciting is getting AI embedded into the smaller, everyday aspects of different clinical workflows in the EHR. Instead of death by a thousand cuts, we can deliver joy from a thousand improvements. There is vast opportunity to go after those smaller opportunities to reduce toil and lower cognitive burden in frontline medical settings.

For clinicians, this starts with going beyond simple scribing tools and ensuring key workflows are deeply embedded and enabled in the EHR—adding structured data to the patient record, referencing existing information, and preparing orders and follow-ups based on the patient conversation. From there, it might look like organizing your inbox, summarizing and predicting what is most relevant in a patient record, suggesting reports you might want to run, prioritizing your task list, making schedule suggestions to spend your time more efficiently. For patients, it could be estimating when you’ll be seen and providing friendly follow-ups.  

For EHR leaders, this means that the “build versus buy” equation for AI requires thoughtful deliberation. Yes, there are strong examples and valuable opportunities for external AI integrations to deliver substantial user value. However, simply layering AI solutions onto existing workflows often falls short of delivering the outcomes clinician users need. Also, as AI opportunities proliferate, front-end users, especially smaller clinics, can’t adopt a thousand solutions to do a thousand jobs. To reshape workflows in AI native ways, some AI competencies need to be built in house. And for tech companies working with EHRs on AI, those integrations need to be more detailed and deliberately implemented into workflows than a cookie cutter approach can provide.

AI is evolving to be not a type of company, but a competency—something that every health technology organization should approach in a way that is deeply thoughtful and curious about applications to specific scenarios and workflows, versus expecting to “paint everything with AI” and profit. We are in a moment of opportunity for the health tech sector to shift the trajectory of progress with innovation that not just saves time, but is closer to “participating” in clinical workflows and patient care—a different type of responsibility that deserves acknowledgement and accountability to the sanctity of the task of improving medicine. 

AI is not a blunt instrument that can be smashed down on any challenge like a sledgehammer. In my experience, out-of-the-box AI solutions can be powerful but, as per usual in healthcare, there are no shortcuts. For AI to really deliver value at the patient, population, and clinician level, it requires a thoughtful, precise approach that is deeply integrated into the tech stack with a scalpel to revolutionize the experience of care from within.