Weekly Health Tech Reads 9/27/26

Today’s newsletter is brought to you by: Elation Health

Open architecture. Real interoperability.

True open architecture integrates seamlessly with registries, HIEs, payer platforms, and care coordination tools — unifying clinical, claims, and operational data across every site and care setting.

If you're interested in partnering with Health Tech Nerds, let us know!

Last week in this newsletter, I dove into some of the developments in the AI personal assistant world, how people are using tools like Instinct, and the implications for healthcare. This week brought another series of fascinating developments to consider in this rapidly evolving world, punctuated by the emergence of the first “platform war” between Amazon and Meta after Amazon blocked Meta’s Muse personal AI agent from shopping on its site.

I’ll come back to that in a bit, but first I wanted to highlight a report I spent time with this week. Citrini Research, the group behind the internet-famous 2028 Global Intelligence Report from earlier this year, penned an update this week exploring the implications of the rise of personal AI agents. This section of the article in particular caught my attention, arguing that friction will move from a profit center to a cost center:

This idea about friction becoming a cost center has particularly interesting implications for healthcare; even the Citrini piece highlights the obvious example of how health insurance has benefited financially from inserting friction into various processes. It seems logical why consumer friction in an AI agent era becomes a cost center for companies — the benefit of friction is that humans give up on a task at some point; AI agents can persist until they get what they want.

In the context of this personal agent explosion, I found Michelle Kittleson’s perspective in NEJM Voices this week to be a great read, as she explores her role as a cardiologist in caring for a patient when opinions differ on a clinical decision. The article started by highlighting a request that one of her patients made via a patient portal:

Kittleson expressed her irritation and frustration at receiving this message from her patient. It’s a good vignette to read from her perspective. Irritation at the way the patient requests this test in a tone like they’re ordering dinner from a waiter. Frustration at the seemingly inevitable result of responding by saying “no” — the patient ghosted her.

We don’t know how the patient felt here, but it is not hard to envision them feeling a similar sense of irritation and frustration at being told “no” and finding another physician who would perform the test. Kittleson concludes by noting that this is a trust deficit and that physicians must work to establish trust with a patient so that saying no is not simply a reason to go find a new physician, but rather a reason to stay as a positive signal that this physician cares enough to say no.

Now, I’d imagine that in this vignette, Kittleson was interacting with another human via these patient portal messages. But imagine a hypothetical where this was actually the patient’s Instinct personal agent reaching out on their behalf, much like we were talking last week about a VC telling their Instinct to get them a dinner reservation. It doesn’t seem that far-fetched that these types of situations are coming, and very quickly.

In that hypothetical, the AI agent sending this request has won the battle that I’m not sure a physician is even aware they’re fighting yet — the agent has the patient's implicit trust to act on their behalf. Saying “no” to the AI agent that has already clearly made up its mind that it needs to order a coronary CT angiogram simply isn’t a realistic option, as the AI agent is just going to continue seeking other options until it finds one, right? The physician saying “no” essentially becomes a gatekeeper, adding friction for the consumer who is seeking a test that they (or AI agent on their behalf) have already decided to get. Going back to the Citrini note, that added friction is just an additional cost, as the AI agent will eventually find a way to overcome it and get the test. The clear financial incentive in this hypothetical world is to say “yes” to those requests; otherwise, they’re just going to go elsewhere.

This is where the Amazon/Meta example, along with the Resy example from last week, invites some really interesting questions about how this world evolves. Clearly, we’re seeing an early signal that companies are going to attempt to block consumer AI tools that they feel are unauthorized. From the consumer perspective, this doesn’t seem to make much sense — think of the VC who was annoyed his Resy account was blocked, posing the question: why is this any different than my human admin booking a reservation for me?

It all points to the complexity of healthcare’s upcoming version of AI agent wars, as healthcare institutions fight over who owns the trusted-advisor relationship at the moment a consumer is deciding what care they think they need. Back to that NEJM article, if you’re not the one convincing that patient whether they need a coronary CT angiogram test in the first place, you’ve already lost them to someone else. That patient is going to get that coronary CT angiogram. Saying “no” may be the right clinical decision, but it’s the wrong financial decision for an organization to make (at least in a fee-for-service world!), and it’s going to be interesting to watch how healthcare organizations grapple with that dynamic, particularly in a market where consumers are potentially controlling more and more of their own healthcare spending.

Personally, I think this is a huge opportunity for trusted healthcare brands to lean in and figure out how to reduce friction for consumers in the world to come. The ones that do seem set to be outsized winners.

Transparent PBM player Rightway announced $155 million in funding this week, at a valuation around $1.75 billion. Francisco Partners led the round, with Thrive Capital and Khosla Ventures both increasing their ownership. Per a Bloomberg article on the funding round, it appears that Rightway has its eyes on an eventual IPO as the endgame in the next few years.

Rightway’s growth is quite impressive, as it reports having 45 Fortune 500 clients as customers. Notably, Eli Lilly moved to Rightway from CVS late last year. This shift seems aligned with the broader market sentiment toward moving away from the Big 3 PBMs to more transparent approaches. Rightway’s success is particularly impressive, as recent data we’ve seen in the PBM market suggests that it is primarily small and medium-sized businesses that have been shifting away from the Big 3 toward transparent PBMs. Recall this chart from the National Alliance of Healthcare Purchaser Coalition survey last month:

The large group market does not appear to have moved yet, so it’s notable to see Rightway’s success in the Fortune 500 market.

One broader nagging question this leaves me with: what is the tipping point at which transparent PBMs transform the industry?

I’ve listened to folks like Mark Cuban rally the general public around the idea that PBMs represent everything wrong with the industry. Per Cuban, the fix is simple: we all need to stop doing business with the Big 3 PBMs, and it will change the entire industry.

Ok, so if that is true, at what point does that change materialize, and what specifically gets better here? Rightway has now captured 10% of the Fortune 500 market, do those employees now feel like their healthcare has been fundamentally transformed? When Eli Lilly shifts employees from CVS to Rightway as the PBM, what do they experience getting better? None of those answers are clear to me. Has anyone reading this newsletter been at a party and had someone come up to them telling them how much better healthcare is with their new transparent PBM? 10% of the Fortune 500 is now using Rightway, so statistically, I’d think we should see evidence of this idea of industry transformation emerging, no?

Don’t get me wrong, I’m all for moving towards more transparent pricing mechanisms in the PBM world. Rightway’s success seems like capitalism at work as employers are moving in favor of transparent models, forcing the Big 3 to adapt or be replaced. That competition is a good thing. It just strikes me that the fairy tale state of industry transformation from transparent PBMs might look much more like a Big 4 in the near future to the average employee experiencing this shift.

Curative, originally a COVID-19 testing provider that pivoted into an employer insurance disruptor offering $0 copay plans, announced that it is removing prior auths from most MRIs. Prior auths will remain in place for a handful of high-cost facilities, and Curative went a step further by naming the 41 hospital facilities that charge between 5x and 20x the median rate for MRIs. Baylor Scott & White and AdventHealth facilities featured heavily on the list. The summary data is below:

As BUCA alternatives like Curative increasingly seem set to enter the mainstream conversation, it will be worth watching how health systems respond to aggressive efforts to steer volume away from them and toward lower-cost settings. I am sure that Baylor Scott & White, AdventHealth, and the others on this list are not pleased, although I doubt their response will include decreasing the prices of these MRIs.

$653 million

BCBS Association’s estimate of incremental reimbursements to hospitals due to AI tools driving up coding intensity

BCBS shared data showing that hospitals are billing inpatient stays as more complex, leveraging AI-assisted RCM tools that are adding secondary diagnoses to claims to increase complexity.

This BCBS data prompted a NY Times piece exploring the payor vs provider bot wars and how these AI tools are driving up healthcare costs. The article highlights McLaren Health System, which is working with the AI RCM vendor SmarterDx to improve documentation of billable codes. This work has increased McLaren's revenue by $1 million per month, of which SmarterDx takes an unspecified percentage. Naturally, there are varying opinions on the merits of this and whether it is a flavor of upcoding or merely more accurate documentation, which unsurprisingly tend to follow payor and provider party lines.

It’s interesting to juxtapose this situation with Citrini's narrative above about friction and the idea that AI agents will turn friction from a profit pool into a cost center. We’re seeing evidence of that thesis in the BCBS Association data, right? Providers now have access to AI bots that are quite good at identifying all sorts of billing opportunities that humans might miss in the RCM process.

All of a sudden, these AI agents appear capable of unwinding the friction from decades of complex payor-provider contract negotiations, as providers have temporarily gained a leg up. The interesting question is now what happens next in these negotiations and how payors respond to the provider shift.

The obvious challenge in this situation is that this is fundamentally a human trust issue. There is a trust deficit between payor leaders and provider leaders who have been engaged in decades-long, often contentious, zero-sum negotiations about how to allocate the healthcare dollar. Check out Andrew Tsang’s recent blog post on the Tufts/Partners negotiation in the early 2000s as a good example of one of these negotiations. This trust deficit drives a Nash Equilibrium in payor-provider negotiations, where each party pursues their self-interest despite an obviously better outcome for both via cooperation. Arming each side of that contentious negotiation with self-interested AI agents only compounds the problem, as the BCBS data above illustrates.

What is the path forward here? Figuring out how to build trust between payors and providers and remove the friction from the equation in the first place. The most logical way I can see to solve this issue actually lies in vertically integrated models, despite what folks in DC appear to think of vertical integration these days. An AI-centric, vertically integrated payvidor model that better aligns the interests of both parties and reduces friction between them seems like a no-brainer. Furthermore, assuming the general idea that friction becomes a cost center holds, the advantage from removing friction should compound quickly as the rest of the market continues to spiral out of control. It bodes well for narratives like Devoted IMO.