Put your healthcare AI app to the test

A call for healthcare AI startups - application link below!

There’s a ton of hype about AI across industries right now, and healthcare and life sciences have been front and center. Perhaps rightfully so – from identifying cryptic binding pockets for drug discovery to automating physician documentation, opportunities to leverage AI across the biomedical field remain vast.

Though AI research has been progressing at a steady pace for decades, with deep learning at the center since AlexNet in 2010, for many, it feels like we’ve experienced a step function improvement very recently. As I wrote about in my piece on Command Line Medicine, innovations in AI, coupled with those in user experience, have exposed the power (and limitations) of new and ever-larger models at scale.

As a result, the capabilities of AI are more tangible to human users than ever before.

Over the last few quarters and in response, healthcare and life sciences board rooms around the world have debated the question “How will generative AI impact our business?” From these discussions, a flurry of new applications have spawned — from Doximity’s DocsGPT product to Veeva’s generative AI features to Epic’s adoption of generative AI for in-basket automation.

In addition to incumbents marketing new capabilities, dozens of startups have emerged (and matured) during the ~160 days since ChatGPT launched. But how many companies are actually solving real, hair-on-fire problems using AI in healthcare and life sciences right now? The answer is fewer than one would think skimming recent headlines.

Today as a medical student rotating through the hospital, I use four AI-enabled applications: Abridge, Atropos Health, Glass Health, and ChatGPT.

• Abridge helps me translate my conversations with patients into downstream documentation (e.g., SOAP notes, patient summaries, etc.) with the push of a button.

• Atropos empowers me to run real-world evidence studies at the point of care, which allows me to contribute to my rounding teams by answering questions that may fall within grey areas of clinical guidelines.

• Glass Health helps me store and track the progression of my medical knowledge and also provides me with real-time, AI-enabled assistance for generating broad differentials and structured plans.

• ChatGPT helps me fill knowledge gaps quickly, such as referencing information about the pathophysiology or presentation of a particular disease.

Aside from these tools, I am not (knowingly) using AI in the hospital day-to-day. I’m using Cerner as an Electronic Health Record, which can’t even flag when a lab value is out of range with consistency. I’m fairly confident that our revenue cycle management department is using some sort of AI to optimize coding, billing, and collections activities that are far abstracted from my work unless they reach out to clarify questions about a chart.

Interestingly, each of the startup products mentioned above is accessible via self-serve and product-led models. I’ve written extensively about why I’m bullish on product-led growth models in healthcare and life sciences (e.g., selling directly to clinicians and scientists via bottoms-up motions), and my own experiences adopting new tools and technology while in medical school have corroborated this thesis.

---

Let’s cut to the chase

Enough pontification. For those of you who are new to my work, I’m both a medical student and bioinformatics researcher at Brown and a venture capitalist with Bessemer Venture Partners. I’ve been investing in healthcare and life sciences for six years and have been focused on the intersection of computation and biomedicine since 2014 through academic research, industry data science roles, and venture capital investing.

I’ve been fortunate to straddle these two worlds, which has granted me a unique perspective on the specific needs of the biomedical industry and how technology, and more specifically AI, can address them.

Far too often in my role as an investor, I meet bright, passionate teams with interesting technology going after problems that don’t really matter in medicine (or at least are not a top 5 problem for clinicians and hospitals). There are many reasons that this happens, though it’s often a mix of techno-solutionist optimism compounded by a lack of unfettered clinical input at the company.

So, I’m making an offer to the AI x biomedicine ecosystem.

If you are building an AI-enabled application where the intended user or beneficiary is a clinician, I would love to try it. In exchange, I will:

• Give you a screen recording of me utilizing the application in the hospital (so long as PHI can be redacted). This will be like working with the UX consultancy The User Is Drunk, except instead of being drunk, I’ll be running on 4 hours of sleep, 3 cold brews deep by 10am with a belly full of whatever stale pastries were leftover from yesterday’s morning report.

• Provide feedback on the product’s strengths and areas of development, as well as my perception of the clinical and financial ROI, drawing upon my experiences as a medical student and also venture investor.

And… if your product provides real value, I’d love to meet with you to discuss your broader vision, product roadmap, and company.

Submit an application here by June 15th at 11:59pm PT. I’ll be selecting five companies for the initial cohort and may consider launching a second at a later date.

If you’ve made it this far, I hope you’ll consider submitting a product, or instead, share this opportunity with someone working on tools that leverage AI in healthcare or life sciences. This should be fun!

Thanks for reading,

Morgan

P.S. If you’re interested in following my journey or future musings about healthcare, life sciences, and AI, subscribe below! Depending on how this exercise goes, I’ll do my best to share back learnings from trying various products on my blog.

---

Thanks to the Thinkboi, Nikhil Krishnan, for his feedback on this project.

Disclosures: I am an investor in Abridge.

A note on safety: As a medical student, I support but do not have the authority to make clinical decisions about patient care. I can, however, share my perspectives about patients and their care after conducting medical interviews and physical exams independently. For any products focused on diagnostics, I will be able to provide feedback on how your technology impacted my workflow and diagnostic process, as well as how my downstream recommendations were received by my team of residents, fellows, and attendings. I will not be able to speak to tasks I am not authorized to perform in my current role.