• [14:45] Healthcare AI adoption has been organic, not mandated like prior health tech eras. Unlike EHRs, interoperability, and telehealth, this wave isn’t being driven by regulation, penalties, or a pandemic. cc: Ambience, Abridge, Bayesian

• [17:00] AI is becoming a labor purchase, not a technology purchase. AI autonomous agents are filling staffing gaps and scaling capacity on demand. cc: Hippocratic, AKASA

• [18:56] AI might *increase* healthcare employment, vs eliminate it. In the healthcare version of Jevons’ paradox, more abundant medical advice and triage creates more demand for human clinicians.

• [26:20] The notion of medical specialties reflects the limits of human cognition. We are starting to see AI-enabled “multi-specialty doctors”, and less risk of patients falling through the cracks because of less need to be passed from doctor to doctor to receive a holistic diagnosis or intervention. cc: Counsel, Firefly

• [28:09] Be AI-native and AI-proof. Infusing AI into the most logistically complex care models that require in-person care, devices, procedures, etc results in some of the most durable business models today. cc: Sprinter, Alchemy, Devoted

• [38:50] Pitch decks now have two audiences: investors and their AI agents. In an AI-screened fundraising funnel, machine-readable detail and specificity is more useful than image-based storytelling.

Listen to full episode here

A top-of-mind question for founders and investors these days is how to build a business that gets stronger as frontier models improve, without becoming vulnerable to them. How do you build a business that is both AI-Native (taking full advantage of frontier models in a way that shows up in the P&L) and AI-Proof (unlikely to be disrupted by frontier models)?

Recent adoption of AI in healthcare has been faster and more organic than in any prior era, and the initial products have largely been focused on 1) better intelligence and 2) automation of administrative work. But healthcare now reflects a broader AI-era theme: as generalist frontier models keep improving, those value props become progressively less defensible, so value moves from intelligence and automation alone to intelligence and automation plus accountability.

Accountability is the Product

Accountability requires capabilities that sit outside the model layer. Even as models improve, they won’t automatically simplify infrastructure and real-world operations, solve regulatory complexity, or facilitate human relationships, let alone make companies more willing and able to bear risk.

When the CEO of a frontier model company wakes up in the morning (if they’ve even slept to begin with!), the priorities on their mind are: lawsuits, staying on the good side of the government, anticipating what China will do, raising enough capital to cement access to compute and top talent, competing with the most formidable companies of all time, and the list goes on… What’s the likelihood that the CEO wants to take on even more regulatory risk and liability by owning the underlying healthcare itself?

Therein lies the opportunity: in vertically integrating into the healthcare accountability layer - where a licensed, contracted, or regulator-approved entity has to deliver an outcome, stand by a claim, and accept the consequences of being wrong (or right).

So which healthcare product form factors own enough of the accountability to be both highly AI-Native and AI-Proof? Here are three:

1. Delivering care and IRL services

   • What’s AI Native: An AI-native clinical services business can operate with a) disruptively low cost structure (e.g. you’re more likely to have run towards the CMS ACCESS program than away from it) and b) faster and more scalable growth and product expansion (e.g. operating nationally, with higher panel leverage, and more specialty coverage, sooner that prior generations of services companies).

   • What’s AI Proof: A medical practice still has to obtain state-by-state licensure, credential with payors, get malpractice coverage, contract and integrate with referral partners, and implement IRL care delivery operations. Even non-regulated, sub-clinical services are subject to rigorous safety validation, clinician oversight, and last-mile integration into care workflows that represent a level of liability that goes far beyond the intelligence layer.
     

2. Bearing financial risk

   • What’s AI Native: AI can arguably automate the majority of the admin work that comes with being a risk-bearing entity, and meaningfully changing the cost structure of managing a population at scale.

   • What’s AI Proof: A full-stack risk-bearing entity may have to raise reserve capital, get state licensure (if fully insured), secure provider network contracts, engage with members and doctors to manage the total cost of care, and win high-touch sales. And even software-only businesses can bear financial risk, e.g. by only getting paid when multi-party transactions are completed or when the customer gets paid.
     

3. Developing an FDA-regulated product

   • What’s AI Native: AI can increasingly automate the operational and cognitive parts of the diagnostic or drug discovery and development process.

   • What’s AI Proof: A company ultimately still has to manufacture the product, put it through years of human clinical trials, and get FDA approval to make a definitive claim about the diagnostic or therapeutic benefit. Not to mention we have only scratched the surface of understanding human biology, so we still need to generate new data and build validation loops at scale to expand the search space over which AI can effectively operate.

Of course, the pace of evolution in the space means this framework could have a shorter half-life than we might think. Deregulation could lower the bar for autonomous AI products to do more without licensure or approval. Agentic provider-network contracting could lower the barriers to fully autonomous health plans in the employer-sponsored market. Frontier models could mature to a point that those existential risks on the minds of their CEOs every morning become far more manageable, freeing them up to explore taking on more risk, liability, and ultimately, accountability.

That said, many of the best-performing healthtech companies today are already implementing one or more of the models above: delivering outcomes, bearing financial risk, and/or developing a regulated diagnostic or therapeutic. And the strongest AI-native performers whose centers of gravity are still “better intelligence” or “administrative automation” are actively figuring out how to take on more accountability - and hence become more AI-proof.

🔥 Healthcare affordability has not improved in the last few years, and has even gotten worse in several areas.

🔥 We’ve also all aged ~20 years in the last 3 due to the rate of policy and regulatory change in gov’t health insurance markets - e.g. MA v28 + new rate notice, Medicaid work requirements under HR-1, ACA subsidy cliff, etc... the health plans are fatigued (and still have a ton of work to do to actually implement and execute against these changes).

🔥 And at the same time - innovation is accelerating! From GLP-1s, cell and gene therapies, to AI-enabled care models. While these technological advancements promise profound health impact, broad affordability for them has not yet been solved. Good policy needs to address market failures related to risk pools, time horizons for underwriting said risk, and pragmatic payment models with teeth behind them.

🔥 Consumer-directed cash-pay is still a rounding error of total spend today... but we should keep an eye on the trend: consumers are waking up to the fact that they’re burning $1,000s inside their own deductible for opaque, overpriced services + a miserable UX. Cash-pay-optimized suppliers are rushing into that void - and people are voting with those deductible dollars. e.g. Function Health, Counsel Health, Truemed, et al are winning a rapidly growing following based on superior value and experience that the insurance-optimized system is not incentivized to deliver. Will these kinds of players eventually constitute the future version of health insurance?

🔥 In 2025, providers were out front on AI while health plans were playing defense... but in 2026, the plans are clearly leaning in, across both internal ops and external member- and provider- facing experience.

In 5 years, if we’re saying healthcare affordability got better, it’ll be because we finally embraced AI to bring down the cost structure of care on BOTH ends of the acuity curve. The low end is obvious. The high end - the sickest, most comorbid patients - is where we were most hesitant to say exactly what the near-term impact could be.... but therein is the grand challenge for builders: bend the cost curve in areas where the system looks the most AI-proof.

Health systems are now among the fastest AI adopters in healthcare. Most are well into “Phase I”, having implemented at least some basic forms of admin automation, patient engagement, and clinical decision support.

The interesting conversations were about what comes next:

• The premise of our Infinite Healthcare thesis: AI Care can flip our industry’s paradigm from “higher utilization is bad” to “higher utilization can be good” because we can deliver it at deflationary prices and preempt deteriorations in health. For health systems, this means being able to serve existing patients 10x better, while also serving entirely new customer bases with new services. The most forward-leaning health systems are doing both.

• A concrete example from Munjal Shah and Dr. Shoeb Sitafalwalla: Hippocratic AI’s “infinite-pilots” (vs co-pilots or auto-pilots) enabled Advocate Health to do things human labor alone never could - like calling every patient lost to follow-up, not just the sickest 10%. When a patient calls to reschedule, the AI also clears any other open tasks while they’re on the line. Clinical and non-clinical issues can be handled through their singular platform - something traditional call centers, segmented by use case, were never able to manage.

• A parallel I discussed with Jennifer Mitzner of Baylor Scott & White: the principles that help a provider succeed in value-based payment look a lot like those of best-in-class AI-native consumer businesses - personalized experiences, proactive engagement, predictive recommendations, profitable membership-based subscription models. Could the highly engaging nature of AI products enable a new performance standard in VBC?

• And with Michael Sanky of Databricks: incumbents have a competitive advantage in the age of AI with their proprietary data assets. But they also have the burden of significant workforce and change management implications of going AI-native. The jobs of doctors and nurses will be unbundled and repackaged, just like the “scopes of practice” of PMs, designers, and engineers are evolving in the tech world.

At the current pace, AI will likely be able to practice basic urgent and primary care autonomously very soon - if not already. So it’s no longer a question of whether the tech can get there - it’s whether regulation can be clarified (federally and across 50 states), whether unions get comfortable, and whether AI-forward payment models can scale and show profit that stands between us and Infinite Healthcare at scale.

What is Infinite Healthcare Worth? As AI becomes a new site of care and automates administrative orchestration, it rapidly eats into $5T+ worth of traditional health care services spend, and potentially represents the biggest market opportunity in all of healthtech.

AI has the promise of bringing the cost structure of care delivery down such that we see much higher ROI per dollar spent. Utilization will likely increase as access becomes more abundant (per Jevons Paradox), and we’ll expect to see entirely new forms of care delivery emerge that we can’t even imagine today. This could be the unlock that finally propels our healthcare system towards a more prevention-centric paradigm, in which we are able to engage consumers regularly and therefore reduce unnecessary, high-cost, and high-risk medical escalations by keeping people healthier to begin with.

In this context, dozens of pricing experiments are already playing out amongst AI care companies, even being facilitated by the federal government (e.g. CMMI's ACCESS program). We lay out a framework for these:

• Per Task

• Per Workflow

• Per Episode

• Per Patient

with Per Outcome as a possible overly on all 4 of those. Companies are blending multiple payment models under roof - sometimes mashing up traditional reimbursement with cash pay, or set fees with outcomes-based carrots/sticks.

One thing that is clear is that we’re in only the very early innings of seeing how AI will transform the topology of the care delivery landscape - how it’s delivered, what kinds of services it will comprise, and how it will be paid for. Core elements of this evolution expressly goes against the grain of the cultural status quo in which increased utilization is viewed as failure, prevention is not incentivized, and cost structure is only expected to inflate, not deflate. But the math of the current system no longer maths - so we must price healthcare accordingly.

Read Our Full Piece Here

• It’s certainly possible now to build a large business purely based on cash pay + DTC acquisition. But playing nice with “The System” (e.g. B2B2C distribution, referrals from trad providers, taking reimbursement risk) still has its merits in helping one achieve even more durable scale and unit economics.

• Sequencing of GTM motions matters: if you go “B2C first”, you have the gifts of being able to fully control your product roadmap and swiftly acquire users - but it will be more expensive and potentially take longer to acquire them on your own. Going “B2B2C first” means you’ll get access to large chunks of users in one fell swoop, but you’ll have to pay a “product tax” to appease your B2B partner’s requirements, not to mention needing to survive through long enterprise sales cycles. This sequencing choice informs your capital raising strategy in the early days.

• A year ago, everyone was scared to say “AI Doctor”, and now everyone is claiming to be building an “AI Doctor”!

• To that end, will healthcare be dominated by an AI Doctor SuperApp as the front door, or will we continue to have fragmentation of apps by use case / condition / demographic? A large portion of consumers who don’t have a PCP will find one in the form of an AI Doctor, and the best AI Doctors WILL be SuperApps that connect into a network of specialists and IRL clinical services through a single front door.

• This necessitates a SuperData layer that allows for context sharing across all apps - and not just trad EHR data, but also wearables, genomics, non-traditional biomarker tests, information gleaned through ongoing engagement with AIs, patient-reported outcomes, health plan benefit design data info, etc.

The buzz in the room felt a lot like the early versions of the Health 2.0 conference in the late 2000’s, BUT with companies that are really working and scaling this time around. Congrats to Zach Teiger and his organizing team for bringing this community together - it’s time to build in consumer health!

Great conversation with Ambience Healthcare CEO Nikhil Buduma about building the infrastructure that’s powering the unprecedented adoption curves we’re seeing in clinical AI, what’s actually hard about clinical intelligence, the future of the EHR, and more.

Watch Here

Listen Here

We led TrueMed’s Series A to support their continued expansion of merchant, payment, and enterprise healthcare partners, as they work to shift the center of gravity of the relationship that consumers have with their health, so that we achieve better balance between prevention, maintenance and treatment.

Read more on why we invested

And the story straight from Justin:

Caroline Merin, Tom Chokel, and Arela Solis are bringing the magic of their platform to physician-patient relationships in 14 countries across 22 medical specialties, and we’re thrilled to be partnered with them on this global journey.

Read more about why we invested

The traditional healthcare system has primarily served three main user segments: (a) “sick MAUs”: people with spiky, high-cost needs; (b) “sick DAUs*”: like those in intensive, long-term care; and (c) “healthy YAUs*”: relatively healthy individuals who rarely see a doctor. Healthy YAUs are at risk of becoming Sick MAUs/DAUs, and preventive care could slow that shift. But our reaction-pilled healthcare reimbursement system rewards treatment over prevention, so access to proactive check-ins and monitoring services are not prioritized, and insurance rarely covers them anyway.

Enter Healthy MAUs: consumers who aren’t actively sick but want to monitor and understand their health on a recurring basis—and also represent the potentially largest portion of the consumer population. We expect a wave of companies—both AI-native upstarts, and repackaged versions of incumbents—to start offering recurring services to serve this user base.

With AI’s potential to reduce the cost structure of care delivery, the advent of novel health insurance products focused on prevention, and consumers more comfortable paying for subscription-based models out of pocket, “healthy MAUs” represent the next high-potential customer segment for healthtech: continuously engaged, data-informed, and prevention oriented.

Read more about this and all the other great Big Ideas for 2026 from the a16z team

* DAU = Daily Active User, MAU = Monthly Active User, YAU = Yearly Active User

Read our full investment thesis

Full Audio; Full Video

Assuming you build a successful product, there will likely be 1 or 2 killer features that your product becomes known for and that deliver the majority of business value (“load-bearing”), and a number of additional features that might help you check the boxes to close deals, engage users, or integrate into enterprise IT environments, but that aren’t really differentiating in and of themselves (“non-structural”).

Thinking back to the products we launched at my own startup (enterprise SaaS for health systems), we probably had 3 key load-bearing modules that we became known for / by which we were distinguished in the market, sprinkled with at least 3 non-structural modules that helped us more credibly position as an enterprise-wide (versus point) solution, integrate into our customers’ IT environments, and serve as a single pane of glass for our users’ operational workflows.

Non-structural features on a standalone basis would probably fall flat on their face, so they have to be evaluated appropriately in the grand scheme of your GTM motion, pricing strategy, competitive landscape, end user workflows, and enterprise buyer criteria. Just because those features don’t have the highest utilization by end users or don’t produce direct financial ROI, doesn’t mean you should throw them out - the psychological, behavioral, and premium pricing value they create in enabling enterprise buyer confidence and long-term user stickiness could be enough justification to keep them in.

From a team comms perspective - it’s best to be clear about which features are load-bearing versus non-structural, and to emphasize that both may be necessary for business success. It can be demoralizing for an engineer to be working on a feature that she thinks is load-bearing but is actually non-structural! Be deliberate about explaining the why behind your prioritization of those features, and try to balance how much responsibility each team has for both load-bearing and non-structural features to have everyone feel like they’re doing their part to maintain the competitiveness of the overall product suite in the market.

Here’s a clip of what he said... and his words are a great reminder of why it’s been so hard to for companies with awesome consumer products to also build awesome consumer health products - an assumption of rational behavior doesn’t work when the underlying system is dysfunctional.

We also discussed:
⚙️ whether it’s possible to move fast and not break things in healthcare
⚙️ CVS as the “everything store” for health and wellness
⚙️ reading into what consumers want for their healthcare needs based on early emergent behaviors with LLM chatbots
⚙️ consumers as a new class of payor, with increasing cash pay rails
⚙️ what the consumer health experience could look like in 10 years

Full conversation here:

These companies are:

• Tearing down the artificial siloes between medical specialties that currently force patients to become their own care coordinators.

• Exponentially scaling clinical judgment through their AI-native care models, enabling effective panel sizes of 10,000+ patients per doctor.

• Implementing multiple business models: adopting everything from tapping into labor budgets, utilizing value-based care, and embracing consumer cash-pay as we’re seeing more broadly with generalist consumer AI products.

• Available on-demand, 24/7, and fully integrated with human care teams.

Some examples of AI-native models include:

• Ubiquitous clinical concierges - every patient will be able to have their own on-demand, clinical concierge, whom they can consult at any time between doctors’ visits: an ally with perfect memory, guidelines-based training, and context-based personalization. The AI concierge can also collaborate with human clinical teams by escalating and assigning tasks.

• Autonomous, dynamic therapy - clinically-trained chatbots will deliver guidelines-based care that’s therapeutically honest rather than reflexively validating. Trained on multiple evidence-based methodologies, they can evolve with each patient’s journey, instead of forcing disruptive therapist changes as often happens in traditional settings.

• Doctor-in-your-pocket “superapps” - AI-native platforms will deliver the best of both worlds: familiar AI chat and voice interfaces, combined with on-demand human doctors who provide clinical-grade diagnoses, prescriptions, and referrals. Purpose-built medical AI infrastructure can unify primary and specialty care functions in a single platform.

These AI-native sites of care are exponentially scaling clinical judgment and expertise - effective patient panels are already expanding from hundreds to tens of thousands per clinician. They're also generating entirely new data sets that are far higher resolution than traditional EHR data - unlocking enhanced personalization, predictive analytics, sophisticated risk modeling, and even better training of future foundation models.

New care models typically demand innovative payment approaches - and we’re already seeing these new AI companies tapping into labor budgets, utilizing value-based care, and embracing consumer cash-pay as we’re seeing more broadly with generalist consumer AI products.

And the scalability and low-cost structure of AI-native sites of care means that we can afford to extend them to even the most vulnerable and underserved populations.

I chose 2011 on the basis of it being the year in which my personal favorite 2010's era PHR / quantified self app, the *original* 100Plus, raised its seed round.

tl;dr - Pretty much all of the 2011 startup categories map to 2025-era activity. Most, if not all, of these categories are looking more viable now than they did 15 years ago for a variety of reasons, e.g.

• infrastructure that didn't exist before

• novel reimbursement rails

• supply-side constraints that have reached a tipping point

• regulatory change that has finally started to play out

• paradigm shift from reactive to proactive care

• step change in consumer demand for preventative health and tools (e.g., GLP-1s)

• consumers as a new class of payor

• greater availability of early-stage capital

• AI / LLMs

So what’s old is new again, but with a confluence of tailwinds that constitute a much stronger “why now.” The categories and mappings are as follows:

There are also several prominent categories that weren’t featured during that era of Health 2.0 that are active now, including RCM, modern health plans, home-based care, and value-based care enablement - these were just less of a thing back then amongst VC-backed players (but were quite active for PE-backed, interestingly).

All this to say - "There are no bad ideas, just bad timing." It’s never been a better time to build in healthcare - this is the industry that will benefit the most from AI, the space is showing signs of evolution that mirror the patterns of earlier eras of the broader tech industry, and traction levels are unprecedented.

Keep at it, builders.

Courtesy of The Wayback Machine, with a shout out to Matt Holt for seeing the future with Health 2.0.

Special thanks to Eva Steinman and Jane Rhee for their contributions.