Is healthcare going to get its own dystopian AI memo?

Anyone who’s been paying attention won’t have missed the viral Citrini Memo. For the uninitiated, it imagines a world where AI has created a “global intelligence crisis” and rocked the economy to its core. According to the memo, which is highly dystopian and intended as a work of science fiction, the job market will dramatically shift in the next few years. And we’re highly unprepared for that outcome. 

The memo is a useful thought exercise, even if many of the experts in my network are skeptical that anything will change all that much in such a short time frame. I’m personally not convinced anyone really knows. Just look at the divisive reaction to tech entrepreneur Jack Dorsey’s decision to layoff half his staff because of AI. I spoke with Zocdoc CEO Oliver Kharraz this week for the Lifers podcast, who predicts that Dorsey will end up hiring a lot of these people back (stay tuned for that episode).

What was useful about the memo is that it did force some major questions about the future of work, and how we can live in harmony with agentic AI as the technology gets more sophisticated. In healthcare, I’m not seeing much thoughtful dialogue out there outside of the usual tropes related to empowerment and augmentation of providers. What about the future of billing codes? Medical education? Care teams? How will unions protect labor? Will AI be deflationary? And what are the winners and losers in that scenario? 

There’s so many important questions that we should be tackling as we consider the future that we are so desperately seeking. Does anyone have predictions of their own? What will change in healthcare by 2028? What won’t change? And why?

Taking us back to the present day, we are continuing to see plenty of news about compounding GLP-1s. Medicare Advantage enrollments are down. And funding is up for AI and healthcare companies. So without further ado, here’s your summary of the news of the week, curated by yours truly, Meredith Nolan (our HBS intern) and our freelancer Annalisa Merelli. 

The news: Faced with an investigation into its compounded GLP-1 as it grapples with unimpressive Q4 results, the company says it’s confident weight loss would be a big business regardless of whether they can offer GLP-1s.

What Hims & Hers said: “We believe there's a really durable weight business, even if you think you're in a kind of draconian scenario of compounding GLP-1s not being there. And I think even more so, when you look into the next year or two, there's an expanding assortment of therapies that I think are going to be very important to patients, and we're going to have to keep evolving that offering in the category just like we do in other categories to make sure that we've got great treatments that patients are really looking for,” the company’s CEO Andrew Dudum told Fierce Healthcare.

The numbers: Stocks of the company fell down 52% year to date, and the outlook on the market went from “buy” to “neutral.”

Why it matters: Federal health regulators have said they’re now ready to rein in compounded versions of weight loss medicines, but per reports, there’s always the risk of another shortage.

The news: A number of private payers have signed a pledge to join the Centers for Medicare & Medicaid Services’ ACCESS model, which covers new technology and virtual care for the management of chronic conditions. 

The reach: An estimated 165 million people will now have access to the model by 2028. 

What conditions are covered? The usual suspects, including high blood pressure, diabetes, chronic musculoskeletal pain, and depression. 

Why it matters: There’s been a lot of chatter in health-tech communities about whether the pricing is too low, so this is a positive signal. This program is designed to be somewhat experimental in nature, given its relationship to CMMI (CMS’s innovation group), so the administration will be watching closely and gathering feedback. 

The news: Medicare Advantage enrollment increased by 1.1 million and reached 35 million people. While the growth continues, it has slowed compared to previous years. 

The numbers: Enrollment increased by 4% in 2025, compared to an average 9% between 2007 and 2024. Humana and Kaiser Permanente were the only large insurers to increase enrollment.

The expectation: Total enrollment is projected to be lower in 2026, coinciding with a reduction in general plans available and an increase of special needs plans.  

Why it matters: Declining enrollment seems to be linked to plans exiting the market across some states, Axios reports. Beneficiaries are switching for the most part because their plan is no longer offered. 

Dive: What will it take for the FDA to approve brain implants? 

Patients with neurodegenerative disorders could benefit tremendously from brain devices that help them talk or move. Those devices exist, but capturing their benefits — let alone design trials and eventually getting FDA approval- is so complex that even the companies furthest ahead in making them haven’t declared the primary endpoints for their trials, O. Rose Broderick reports in a fascinating feature for STAT.

NVIDIA’s latest report captures the extent to which the healthcare industry has scaled up the use of AI in the past year. Among the organizations surveyed for the report:

80% saw reduced operational costs due to AI

47% is considering agentic AI

61% medical imaging organizations are adopting AI

57% pharma organizations use AI for drug discovery

85% plan to increase their AI budget

The results suggest ChatGPT Health has some room to improve as millions of consumers start using AI agents for triage, especially when it comes to behavioral health. But experts say the study had no control group. Is it fair to compare ChatGPT Health to perfect conditions that are so rare in the real world? Our report here.

The Peterson Health Technology Institute, which assesses whether health-tech vendors that sell into employers and health plans work, has previously tackled the diabetes and MSK segments. Its latest analysis focuses on virtual GI care. Read it here.

Heidi Health convened a panel of medical experts globally to talk about the issues we’re facing today, inclusive of supply shortages, variations in physician quality, and the growing affordability crisis. The teaser video for the full series is here:

$6.5 million for Baba: The company emerged with a seed fund led by General Catalyst to develop its platform for advocacy of Medicare-covered older adults. 

$45 million for Ten63 Therapeutics: The AI discovery company announced a new investment that brought its total funding to $45 million. 

$35 million for Codexo: The pre-claim payment integrity platform, which works by intervening before claims exist through AI education, closed a Series C round led by CVS Health Ventures.

$25 million for Turbine: The virtual drug discovery company closed a Series B round led by Interactive Ventures Partners, and announced a partnership with 10 pharmaceutical companies. 

€ 20 million ($23.6 million) for Nyra Health: The AI-based neurorehab platform closed a Series A round to expand its platform across Europe and the US. 

Flinn AI raises € 20 million ($23.6 million), too: The startup that uses AI to automate compliance for medical device companies closed a Series A round. 

$15 million for Rainfall Health: The compliance and reimbursement AI platform for hospitals closed a Series A round led by Two Bear Capital.

$14 million for Handl Health: The health plan design technology company closed a Series A round led by Arthur Ventures.

A $3 billion valuation for Grow Therapy: Grow Therapy, an all-in-one platform for therapists, is getting rewarded by investors for achieving that all too rare combination of profitability and growth. 

Cristiano Ronaldo invests $7.5 million in Herbalife: The billionaire footballer acquired 10% stakes of Herbalife’s HBL Pro2Col Software, a wellness platform providing personalized health protocol.

Hims & Hers acquires Eucalyptus: The acquisition deal of the Australian digital health company is valued at $.1.15 billion, of which $240 million is cash, payable at the completion of the deal.

Epicured acquires Chiyo: The healthy food delivery service acquired the women-specific nutrition platform offering specific dietary plans tailored to the stage of a woman’s life. The terms of the deal have not been disclosed. 

…to market your next role! Second Opinion will be building an expanded jobs board product alongside our collaborators Aequitas Partners (curator of an already very popular industry job board!). It includes:

With our combined reach, jobs will be shared with tens of thousands of executives within the broader healthcare technology ecosystem across commercial talent, operational leaders, engineering, marketing, clinical, and more. Stay tuned or reach out for more information!

SO: I was on a panel recently with a group of physicians talking about how they use AI. And one of them made the point that there’s a perception that using AI is a bit of a crutch. Someone else shot back and said there’s something wrong with a physician who refuses to use AI given at your fingertips, you have a vast corpus of medical knowledge in seconds. And this debate ensued. Where do you fall on the spectrum? 

BW: I think the former point is ridiculous. Is my stethoscope a crutch? Is my CT scanner a crutch? My job is not to keep my life as intellectually interesting as possible. My job is to take care of patients and make them better, and try to do it at the lowest possible cost and friction. Any tool we can use to do that is positive. That said, I do understand the argument that we’ll de-skill or become stupid, and that is problematic… But before we say deskilling is bad, I rely on my phone for my memory in many cases, but it’s okay because I have my phone. It’s important for those of us in the medical education business [to think through this, especially]...If people use it as a crutch and don’t learn clinical medicine or diagnostic reasoning. And if we count on them to oversee the results of imperfect AI, then we have a big problem. 

In the period where it’s good enough to be useful but not perfect, the idea of de-skilling is really quite problematic and no different, really, when I think about my life as a writer. Do I become a worse writer if I rely upon Claude? Probably. But we need to keep our eye on the ball. And the ball is what tools can we use to make care better, safer, and less expensive?

CF: So Bob, one of the things that I loved about the previous book - The Digital Doctor - you had a second sub-line of the book, which explored hope, hype, and harm. This book, I thought, was really good at providing a balanced assessment of what really was the first generation of a lot of these health technology tools, whether they be the EHRs, the early wearables, some of these digital health companies. 

Then, with your new book, you talk about the idea of the giant leap. So you've gone from quite a balanced assessment and at times leaning quite skeptical about the role of technology in health care to now seeing this massive potential. But there’s the technology. And then in the world of reimbursement, this fee-for-service world that we live in. How do we make sense of these incredible technologies, the efficiency gains, the productivity gains we could get from them, while also fitting them into a world that was not designed for AI?

BW: Yeah, I say first off, Chrissy, when I wrote The Digital Doctor, it was because I was really pissed. I mean, it was like, why is this technology, which we've been waiting for 30 years, so disappointing? And why is every doctor I know unhappy about it?

And so I wrote, it really is, the main character is the electronic health record and why things have gone so badly. And one of the take-home messages was that unanticipated consequences are to be anticipated. This is harder than it looks.

And much of the value we get out of these tools is going to be through implementation and in the context of the policy environments. You're actually right to pay a lot of attention to that. When I started writing this book and I ended up doing about 110 interviews, I was not sure where I was going to land, whether I was going to land in an optimistic place or a neutral place or a pessimistic place.

I landed in a very optimistic place in part because the needs are so great, in part because the tools are so remarkable, and in part because, yeah, it doesn't do anything fundamentally to change the policy environment. But the policy environment is the one we have today. I mean, it's an environment that is in many ways counterproductive, in many ways drives overuse.

I can see how these tools will insert themselves and, even in that environment, make things better. Now, you may argue that in the current policy environment, we won't use these tools because there may be no incentive to use them if what they're going to do is make care less expensive. I think that could go in a lot of different ways, but to me, if we say part of the big challenge in health care is the cost of care, $5 trillion a year in the US and 20% of the GDP, which obviously is true.

I could see the tools making care less expensive, but it's not a slam dunk. The two ways it could make care less expensive are by lowering administrative costs and leading to far fewer FTEs in our health care delivery environment. I could see that happening and it's already going to happen.

We have 1,000 people in our billing department at UCSF. I don't think we need 1,000 FTEs to do billing. The second way it could make care less expensive is through clinical decision support and guiding me as a doctor to deliver the care that actually is the best care at the lowest cost.”

Will it do that? Health care is pretty good at figuring out ways of not cutting our costs and actually raising our profits. We all operate on pretty slim profit margins, so will it guide me to the least expensive care, or will it guide me to the care that delivers the best reimbursement for my system?

I think that's a decision policymakers have to make as they're sort of grappling with this technology, but it will give us the capacity to try to influence the care that's delivered, allow patients to do certain things themselves or with lower-priced professionals than seeing me. I think in some ways this is all in our hands and we've got to figure out what the policy envelope is to wrap this in. But it's hard to believe that having these kinds of tools that can allow us to use evidence in a new way, to deliver care in a new way, and to get rid of so much of the bureaucratic silliness and so many of the FTEs that we just hire and hire and hire and we can't afford and we can't even find them anymore.

It's hard for me to believe that's not net good, and that's partly how I landed in a pretty optimistic place

CF: Is any part of you reacting with fear? Look at what these technologies can do. Is the job going to fundamentally change? We don't know how it's going to change. What about replacing certain medical specialties? How scared should physicians be, or should they instead be kind of shouting from the front lines that they want to use as much AI as possible?

BW: I get this question all the time from people: Should my kid become a doctor? If doctors are gone, it means that journalists are gone, accountants are gone, and lawyers are gone. Writers are gone.

All those things happened before, doctors, because the stakes are high. The regulatory environment is such that I don't think people really want a bot to tell them that they have cancer anytime soon. So I think we'll probably be the last to go, but I think we are all wrapped up in that big question of job replacement, which I think is one of the fundamental questions that this raises.

CF: Bob, reflecting on just the process of this book, what were some of the highlights, lowlights? Was this one easy to write?

BW: A joy. I have to say, first of all, I like to write because I like to think, and I like to try to, you know, part of what I do for a living is try to think about big, complex, multi-dimensional issues and then figure out how to make sense of them for both people in my profession and people outside of my profession. I did a lot of that during COVID, and I enjoy doing it.

I enjoy taking in the information and learning from people. I'm very curious. And then seeing if I can make sense of it myself, and then translate it. When I was asked by the publisher to think about writing a book on healthcare AI, which was exactly two years ago, I said no. And I said no for two reasons. First of all, there's going to be a ton of books out there already.

And the answer is there are some, but none that do what I think I've tried to do, and I think successfully. And the publisher kind of convinced me of that. Because a lot of these books are written by tech experts, and I don't see myself as a tech expert.

I see myself as an expert in the system and how it works, and how doctors think and how systems kind of organize themselves… The second reason I was reluctant to do it is I said, the minute this is out, it's going to be out of date. And my publisher, to their credit, said, if that's true, then you've written the wrong book. I think that the book I tried to write was one that's not dependent, are we on GPT-5 or GPT-12? But ask the big fundamental questions, exactly the ones you're asking.