Trade-offs and transparency
This post raises more questions than answers about trade-offs in healthcare and its business models. There are no simple answers…and certainly they cannot be expressed in numbers.
VS
This week, I have been mulling over all the issues around peptide mills on the one hand, and the AI and business models offered by companies like Onix (Onix.life) on the other. Onix is a startup AI platform that has enabled a roster of 17 (and growing) physicians and other health professionals to each transform their unique, extensive knowledge into carefully curated and vetted avatars (text only). Their business model shows how carefully used AI can actually provide useful, personalized information at scale - but only when it’s carefully constrained. The trick is to start small, with a limited corpus, rather than to start large and try to control things with guardrails that are inevitably complex and focused on removing dangers that are hard to specify or anticipate.
Last month I attended Onix’s New York launch event, showcasing three of its human experts along with an audience-led conversation/texting session with the avatars. Onix calls them onixes (or onices? asks the pedant in me) or “experts,” but specifically not “doctors.” Techies would call them Small Language Models (SLMs): They have the language skills of a standard LLM chatbot, but each one is focused and vetted to include only a “small” subset of information approved and continually augmented by each human expert.
From her side, Ashley Koff RD, one of those experts, says: “I joined Onix because my patients are asking other AIs for my professional opinion, and those AIs were pretending to be me. The ability to curate and train my own AI, solely from my approved content and practice methodology honed over 25 years, felt meaningful to offer as a competitive and better choice. We [at Onix] are longstanding experts with practical application insights to share, versus open internet sources which are the exclusive source of [most] other AIs. I’ve spent hours each week for over 6 months and continue to do so to train and add content to my onix.” That time, spent and continuing, will enable Koff and her colleagues to serve many times more people - with personalized advice - than she could ever reach personally.
The result is a genuine benefit to and model for society, without the rampant dangers of sycophantic bots or misleading “AI companions.” Users get all the time they need with an AI expert that can provide advice specifically tailored to their individual situations - as they report them in response to the AI’s questions. Some will use it wisely, listing all the specific factors in their particular situation that a good doctor would want to know about but rarely has time to probe for; others might expect instant answers. Fortunately, the AI expert is pretty good at pressing for details, and doesn’t have other clients - or duties - waiting. For the users, it’s amazing: They can get good advice and insights, specific to their own case: all their past conditions and experiences, their diet, their daily rhythm. In short, it allows clients to ask detailed questions - and get patient answers. Says Nicholas Nadeau, co-founder and CTO: “Our concrete guardrails are no diagnosis, no prescribing, scope limited to what the human expert is expert in, only the expert’s own approved material as a source, and escalation to the person’s own provider for anything clinical.”
The users’ data sits on their own phones and on infrastructure Onix controls, in Canadian and EU regions where possible, independently audited for SOC 2 Type II and HIPAA, with as much as possible kept encrypted on the user’s own device. To quote the Onix website FAQ: “Onix is based in Montreal under North America’s strongest privacy laws.”
The business models are aligned with simplicity: Onix and its human experts share the subscription revenue that the AI experts will generate. There’s the rub: Not everyone can afford such a subscription. When I prompted Nadeau with “Have you thought about Tom’s Shoes?” I was happy to hear him respond, “Yes, we’ve been considering some kind of buy-one, give-one.” There’s also the possibility of working with public payers and insurers as possible routes to affordability.
Meanwhile, Onix and its AIs do not sell drugs or supplements or devices. Its revenue-share model is simple, though it will inevitably get more complex as more subscribers use the multi-onix mode. The medical professionals, including the three present at the soiree (Ashley Koff, RD; Jordan Metzl, MD; David Rabin, MD), get a revenue share and visibility, plus the opportunity to mention books (for some of them) as appropriate. Says Nicholas Nadeau, co-founder and CTO: “Experts keep 70 percent [of revenue generated through subscriptions to their onix]. They set their own price. It is non-exclusive; they can leave any time and we delete their model and the data [while the subscriber can keep their own text file of the conversations]. If the experts do not make money, we do not make money. They are not our clients, they are our business partners.” The rest of the human experts’ professional lives - whether selling books and medical services or recommending prescriptions on behalf of a clinic or hospital - is separate from their relationship with Onix.
But business, just like the intricacies of human health, is complicated and changeable. The DNA of mission inevitably confronts the metabolism of money. It is worth considering the questions and the trade-offs and business models more closely, for Onix and for healthcare more broadly. Just last week the peptide lobby - sorry, but that’s the best way to describe them - was inserted by the FDA’s parent agency, the Department of Health and Human Services, into an FDA panel hearing on the sale of unapproved (not quite disapproved) peptides, and as of Friday that panel gave their approval to 6 out of 7 of the proposed peptides; now it’s up to the FDA itself to make a final decision. The FDA staff are against the move, but HHS leadership (RFK et al.) in general supports it.
Some of the pro-side panelists and their allies argue that FDA approval would actually mean more vetting of the suppliers and the purity of the drugs, versus the situation now where the FDA simply says “Go ahead; we are not regulating these so you are free to sell them.” But that’s a hollow promise. Just for context about what’s really at stake here for the sellers, a recent article in Barron’s (the Wall Street Journal’s even more market-focused sibling publication) quotes an analyst predicting that non-GLP peptides could quickly amount to a $30+-billion market.
Of course, the entities selling peptides (GLPs and others) have a completely different business model from that of Onix; all Onix is selling is subscriptions, and all it is sharing with its experts is those revenues. Meanwhile, if the FDA does approve the peptides I mentioned above (or others), we’ll see a further onslaught of the sort of AI-driven pill mills you can read about in WIRED, and likely again elsewhere as the FDA drama progresses. This market depends on the original sin of product sales, which unfortunately are more profitable than the best, most tailored advice in the world.
The actual experience
So, that’s the set-up. How is the actual onix experience? I chose to talk with Ashley Koff, RD, both onix and human, because I’m curious about the long-term impact of GLP-1s: Clearly, they help with diabetes and obesity, and clearly they have alarming side-effects for some people. They also seem to help with cravings for things other than food - not surprising, given that they work on receptors in the brain as well as the gut. Onix is still in beta, so I worked directly with the company for access - both as a reporter and a prospective investor. It was an interesting experience.
I started with the onix “expert” Ashley, the “Weight Health Nutritionist,” leading with some questions about electrolytes (which human Ashley wrote about recently on her Substack), and then the issues around GLP-1s…perhaps too much all at once. How much do GLP-1s provoke your body into healing itself, vs. replace something you’re missing? It/she jumped into the electrolytes quickly: I had heard of them but never paid much attention to what I considered a sports-dude issue. Nor had my PCP ever mentioned them. My PCP *did* tell me a while ago that I was dehydrated and should drink more water. Onix Ashley gave me a brief tutorial, and then asked me about my diet. Then it suggested that I take a water test - drink 8 ounces of water and see how long before I needed to pee - so that it could give me better advice.
Though I tried to move on to the broader questions about GLP-1s, the onix kept politely sidestepping my musings about the difference between prevention and treatment, and how GLP-1s fit in. It quickly turned the conversation from a philosophical examination back into one about my own “situation,” structured more like a traditional medical discussion. This focus on the specifics of the individual’s circumstances is by design: mass personalization at scale. Talking to the real Ashley a day later was much more satisfying…. But then, I’m a reporter, not a client (at least not for now).
Nonetheless, onix Ashley did a great job. It kept listening and collecting facts for longer than any doctor has time for…and honestly, longer than I had time for! (No other clients in the queue waiting for me to finish, already!) I could watch it building a simulation of me just as complex as the app’s simulation of the real Ashley. The medical information it provided was interesting but incredibly complicated; had I actually needed it, I would have been grateful to have the onix sort it all out for me.
Net net, I got the unsurprising confirmation that one of our biggest healthcare problems is that we so often treat the body as a collection of body parts. Got a problem; apply the solution! But the body is actually an extremely complex and integrated network of interacting and interconnected systems. For example, GLP-1s bind to receptors not just in the stomach and gut, but also in the pancreas and the brain. And just think for a moment about all the warnings you get on the label of pretty much any drug, from aspirin to penicillin, about what it may do in addition to what you are taking it for.
A day later, over the phone, the real Ashley Koff described her own approach: “People keep talking about tools as if they are about solutions. But there are few solutions in healthcare. I have a toolbox, and I need to know enough about my patients to know which tools to apply where. Then we gather feedback from how the tool performs to determine how we move forward. It’s ongoing; it’s optimization, not diagnosis and prescription. People are taking GLP-1s for all kinds of reasons, and with all kinds of underlying conditions. Yet the standard pharma approach if it’s not working is just, ‘‘Increase the dosage.’”
On the other side, you have the promise of Onix and its onixes. Koff herself is out of most people’s financial reach. But her insights are not. For now, anyone can read her Substack for free, and people like me, with a pre-launch code, can converse with onix Ashley at length. Soon, they’ll be able to subscribe. Pricing has not been announced, but Nadeau says the target is much less than what a single appointment with the expert would cost, if you could even get one. Onix Ashley may not be “real,” but in so many ways she’s better than your average doctor - and has a lot more time than Ashley Koff, RD, herself.
That’s the advice part. But to actually get any medications onix Ashley mentions, you need to go through a pill mill or a legitimate doctor, who may still press you to up your dosage just to move things faster, whether for their own business reasons or because they are eager for “results” - despite the possibility of side effects.
In this real world, I’ll take onix Ashley plus what I hope will be a growing cadre of good doctors who feel comfortable relying on services like Onix to supplement their own time and expertise. People are in desperate need of personalized attention and advice as they seek medical care, and I hear again and again that they are not getting it. Many PCPs now have little time to do anything but prescribe and refer.
Meanwhile, at the soiree, Onix demonstrated something that is not yet part of their formal offering: Conversations about and with a patient among the three professional experts present. That’s what so many patients long for, and hardly ever get. Why can’t my PT guy actually talk to the surgeon? Or could the endocrinologist who manages my bone density please consult with my dietitian?
That’s an enticing possibility. In the long run, the kind of trust Onix can earn is too valuable to throw away. The very messiness of the healthcare market makes their example more important - and more notable. I will probably throw in some money and then fight like crazy to help the company stick to its mission. Meanwhile, I hope Onix’s lead also encourages some other honorable companies to join the fray. Healthcare needs more trust - and that will happen only with more trustworthy players.
As always, across so many domains, the potential of the tool rests in our ability to know when to tap it to extend our own abilities and how to feel confident in that decision, and the danger lies in relying on it without fully applying ourselves to understanding it.
Thank you Real Ashley :) @Ashley Koff RD