A16z's Yoo Says Software Deficit Gives Healthcare Agentic AI Advantage

Your AI Doctor Is Coming | Julie Yoo — Julie Yoo, Sophia Dew

Julie Yoo, Sophia Dew

Healthcare's historic delay in adopting enterprise middleware spares it from massive software replacement costs, allowing providers to leapfrog directly into agentic AI that cuts service delivery costs up to a hundredfold.

AI-assisted summary of Your AI Doctor Is Coming | Julie Yoo — original source. Check the original for full context.

Source published:

Key ideas

  • Healthcare's historically low IT spend and lack of middle-tier enterprise SaaS allow it to leapfrog directly from basic electronic health records into agentic AI workflows, avoiding costly rip-and-replace cycles. (10:00)
  • Unlike previous top-down mandates such as Meaningful Use or pandemic emergency rules, the generative AI wave marks healthcare's first organic adoption trend, driven by clinicians enthusiastically pulling AI scribes to cut charting time. (15:00)
  • AI models have collapsed the unit economics of delivering credible medical advice by roughly 100x, fueling a direct-to-consumer healthcare boom where patients pay cash out-of-pocket for transparent, low-cost subscriptions. (20:00)
  • Within ten years, continuous personalized data collection beyond episodic yearly clinic visits is projected to power a dedicated, lifelong AI doctor in every individual's pocket. (25:00)

Detailed notes

## Background Published on September 6, 2026, on RSS/web feeds by Julie Yoo and Sophia Dew, this 27-minute episode features Andreessen Horowitz General Partner Julie Yoo discussing how healthcare could benefit more from AI than almost any other sector. According to the show description, Yoo traces the industry's shift from paper charts and fax machines to modern digital platforms, explaining why historically slow tech adoption allows medicine to leapfrog legacy SaaS directly into agentic AI, lower care delivery costs, and enable lifelong personalized AI doctors.

## TL;DR In this discussion, a16z General Partner Julie Yoo unpacks why healthcare's legacy underinvestment in software creates an unexpected leapfrog advantage, allowing the sector to bypass decades of SaaS middleware and adopt agentic AI directly. Driven by clinician burnout, rising consumer out-of-pocket costs, and modern foundation models, medicine is witnessing its first purely organic adoption wave through tools like AI scribes and conversational triage. Yoo outlines three distinct frontiers—democratized intelligence, verified clinical action, and longitudinal care—and highlights massive founder opportunities in direct-to-consumer cash-pay services and full-stack, AI-native clinics. Ultimately, she predicts that continuous physiological data collection will shift medicine toward hyper-personalized 'N of 1' care, giving every individual a lifelong AI doctor in their pocket.

## Atmosphere & style The conversation carries an energetic, tech-optimistic tone characterized by rich insider anecdotes from Julie Yoo’s 15-year journey building Kyruus and investing at a16z. Rather than a dry academic discourse, the dialogue moves briskly with an approachable, collegiate rapport between host Sophia Dew and Yoo, who blends pragmatic business realism with palpable founder envy. It feels like an authentic, forward-looking strategy brief where bureaucratic frustration gives way to genuine excitement about human-centric technological renewal.

## Bulletin - `10:00` Healthcare's historically low IT spend and lack of middle-tier enterprise SaaS allow it to leapfrog directly from basic electronic health records into agentic AI workflows, avoiding costly rip-and-replace cycles. - `15:00` Unlike previous top-down mandates such as Meaningful Use or pandemic emergency rules, the generative AI wave marks healthcare's first organic adoption trend, driven by clinicians enthusiastically pulling AI scribes to cut charting time. - `20:00` AI models have collapsed the unit economics of delivering credible medical advice by roughly 100x, fueling a direct-to-consumer healthcare boom where patients pay cash out-of-pocket for transparent, low-cost subscriptions. - `25:00` Within ten years, continuous personalized data collection beyond episodic yearly clinic visits is projected to power a dedicated, lifelong AI doctor in every individual's pocket.

## Key takeaways - `00:00` The transition to electronic health records (EHRs) was driven by massive federal subsidies paying doctors tens of thousands of dollars to abandon paper filing cabinets. - `02:27` Long patient wait times in metropolitan areas were revealed to be an operational routing failure rather than an absolute shortage, as physician scheduling capacity hovered well below 100%. - `05:00` Building health tech in 2010 lacked modern cloud, API, and billing infrastructure, requiring founders to interface directly with on-premises hospital servers built in the 1970s. - `05:00` COVID-19 disrupted traditional regulatory orthodoxies by temporarily suspending state-by-state physician licensure barriers and establishing telehealth reimbursement parity. - `06:48` A convergence of consumer expectations set by on-demand apps and severe institutional margin pressures has pushed healthcare incumbents to rapidly adopt automation. - `10:00` High-deductible health plans exposed consumers to upfront out-of-pocket costs, catalyzing demand for transparent direct-to-consumer alternatives. - `10:00` Clinical expertise requires at least seven years of education, creating a rigid labor bottleneck that software could not meaningfully augment until large language models emerged. - `15:00` Millions of patients now use generalist LLMs like ChatGPT and Claude for rapid triage, bypassing months-long specialist delays and expensive emergency room visits. - `17:30` AI healthcare spans three buckets: access to intelligence, clinical verification and execution, and longitudinal care across a patient's life. - `17:30` While foundation models solve the intelligence bucket, the highest-value venture opportunity lies in solving regulated last-mile actions and continuous lifelong care. - `20:00` Venture attitudes toward consumer health have inverted, transforming out-of-pocket cash-pay products into some of the fastest-growing startups in the sector. - `20:00` Winning health tech companies are building full-stack models that look like physical clinics on the outside but operate on automated, AI-native infrastructure on the inside. - `25:00` Current clinical machine learning models are fundamentally constrained because episodic yearly clinic visits fail to capture the continuous daily physiological narrative of human life.

## Detailed walkthrough ### `00:00` From Code to Cabinets: Digitizing Medicine and the Origins of Health Tech

Healthcare remains one of the costliest industries on the planet, plagued by acute shortages of medical expertise and provider time. Introducing the episode, host Sophia Dew and Andreessen Horowitz (a16z) General Partner Julie Yoo set out the premise: healthcare’s historic sluggishness in adopting software might paradoxically become its greatest competitive advantage, enabling it to bypass decades of clunky enterprise middleware and jump straight into AI-native workflows. Sophia highlights that while artificial intelligence discussions often feel abstract, medicine offers the clearest, most visceral demonstration of how AI directly impacts human lives, prompting Julie to recount her unexpected entry into the field eighteen years ago.

Julie originally trained as a pure technologist, working as a software engineer for the first seven years of her career building early-era e-commerce, fintech, and manufacturing platforms. Looking across the enterprise landscape, she saw modern software transforming nearly every major sector, while healthcare remained mired in physical paperwork. Julie recalls that anyone engaging with a doctor’s office was being “papered to death,” surrounded by towering filing cabinets filled with manila folders. A turning point arrived when the federal government stepped in with a massive financial carrot: paying clinicians tens of thousands of dollars to adopt digital systems of record. Julie realized that the timing was right to translate the enterprise software lessons she had learned in tech into the medical sector.

“Basically, what they said was we will pay doctors tens of thousands of dollars to implement electronic health records,” Julie Yoo explains. “And that was the way by which our industry sort of came into the modern era was for over the course of about five to seven years, I would say. Literally 100% of doctors in the U.S. were heads down implementing these kind of system of record, electronic health record platforms.”

- **The Federal Incentive Push:** The federal government’s multi-billion-dollar program subsidized physicians with tens of thousands of dollars per practice to deploy electronic health records (EHRs), transitioning the entire American clinical workforce away from paper charts in a five-to-seven-year sprint. - **Enterprise Precedents:** Julie’s pivot was sparked by watching enterprise technology reach maturity in retail, manufacturing, and financial services, leaving healthcare as the last major frontier ready for digital modernization.

---

`02:27` Computational Biology, Kyruus, and the Patient Access Paradox

Julie’s initial foray into health tech began on the computational biology frontier during the wake of the Human Genome Project. For the first time, scientists were deciphering the core code of human biology, giving rise to an explosion of biological data and startups linking genomics directly to clinical disease mechanisms. While deeply interested in data-intensive biology, Julie soon encountered a far more mundane yet pervasive structural bottleneck: basic operational access. Across every major American metropolitan area, patients consistently faced wait times stretching weeks or even months just to secure a standard primary care appointment.

When Julie and her co-founder sat down with chief executives and administrators of large health systems and multi-specialty physician groups, they uncovered a startling reality. While waiting rooms were overflowing and patient queues stretched out the door, the actual capacity utilization of their physicians hovered far below 100 percent. Recognizing this disconnect as a massive logistical routing failure rather than an absolute supply shortage, Julie co-founded Kyruus in 2010, serving for nine years as its Chief Product Officer. Kyruus built a unified inventory and scheduling engine designed to intelligently route patients to the appropriate specialist and time slot, optimizing clinical capacity across fragmented hospital systems.

Building in health tech in 2010, however, felt like operating in the technological Stone Age. The modern plumbing developers take for granted today—ubiquitous cloud infrastructure, developer-friendly developer tools, and modular API layers like Stripe—did not exist in healthcare. Mainstream venture capital largely avoided the sector, treating medicine as a regulatory quagmire where enterprise software sales went to die.

```mermaid flowchart LR accTitle: The Patient Access Paradox accDescr: A diagram contrasting patient demand bottlenecks with underutilized physician schedules. subgraph DemandSide ["Patient Reality"] A["Patient Seeks Care"] --> B["Weeks or Months\nof Wait Time"] end subgraph Bottleneck ["The Missing Routing Layer"] B -.-> C["Kyruus System of\nAppointment Inventory"] end subgraph SupplySide ["Health System Reality"] C -.-> D["Real-Time Capacity\n& Specialty Matching"] D --> E["Physician Utilization\nWay Below 100%"] end ```

- **The Patient Access Paradox:** Despite urban patients waiting months for routine primary care visits, clinical scheduling audits revealed that provider capacity was routinely and substantially underutilized due to fragmented booking workflows. - **Kyruus Core Thesis:** Founded in 2010, the company built a real-time, searchable appointment inventory layer that routed patients to the right provider, matching clinical needs against available scheduling blocks. - **The Stone Age Environment:** Health tech startups in 2010 operated without standardized APIs or digital billing rails, navigating a risk-averse climate with minimal institutional venture backing.

---

`05:00` Chipping Away at Concrete, the COVID Catalyst, and the AI Influx

Reflecting on raising capital in the early 2010s, Julie jokes that only about five venture capitalists were actively writing checks in digital health, and she ended up raising from every single one of them. Enterprise integrations meant interfacing directly with legacy on-premises hospital servers built in the 1970s. Top-tier engineering talent routinely chose Silicon Valley consumer tech giants over healthcare, viewing the medical sector as bureaucratic, slow, and thoroughly unglamorous. Despite those head早期 headwinds, Julie’s persistence culminated in Kyruus being successfully acquired the prior year, marking what she terms a “15-year overnight success.”

Stepping back from operational roles to lead the healthcare investing practice at Andreessen Horowitz, Julie observed two profound macro tailwinds that completely reshaped the landscape. The first was the COVID-19 pandemic, which shattered longstanding regulatory and operational orthodoxies overnight. Historically, physician supply had been intentionally constrained by state-by-state licensure laws, preventing out-of-state clinicians from practicing across state lines. Concurrently, commercial and government payers had maintained strict reimbursement rules that refused to pay for telehealth visits unless a prior in-person visit had occurred. When physical clinics shut down, state licensing boards and federal agencies were forced to relax constraints, creating a truly national demand model for remote clinical services and proving that care delivery could occur beyond clinic walls.

The second, even larger inflection point has been the rapid emergence of generative and agentic AI. Julie argues that healthcare is uniquely positioned to benefit from AI more than any other sector because of its deep operational inefficiencies, massive unstructured documentation burdens, and systemic labor shortages.

“I liken the work that I did literally with my own two hands as taking a chisel and a nail against this concrete wall of integrating into the infrastructure of health care,” Julie Yoo explains. “I mean, I think I've declared loudly that I believe healthcare is the industry that will benefit the most from AI for many reasons that I'm sure we'll talk about.”

- **The 1970s Infrastructure Barrier:** Early digital health engineering required building bespoke adapters into hospital software stacks dating back to the 1970s, making talent acquisition and technical execution exceptionally labor-intensive. - **Licensure and Payment Deregulation:** Emergency pandemic policies suspended state-level licensing friction and forced insurers to reimburse virtual visits at parity, creating nationwide distribution channels for digital health providers. - **The a16z Healthcare Pivot:** After orchestrating Kyruus’s exit, Julie transitioned to lead healthcare investments at Andreessen Horowitz, tracking a three-to-five-year acceleration of adoption curves fueled by modern foundation models.

---

`06:48` The Healthcare Renaissance: The Consumer Gap and Incumbents Under Siege

Sophia observes that the cultural standing of health tech has dramatically reversed. Rather than avoiding medicine in favor of frictionless consumer apps, premier AI researchers and founders—such as Anthropic’s Dario Amodei publicly declaring commitments to curing diseases—are aggressively targeting healthcare's most complex, bureaucratic frontiers. Sophia asks Julie to isolate the exact structural shift that ignited this ongoing industry renaissance over the past few years.

Julie attributes the shift to a convergence of consumer revolt and institutional economic desperation. On the demand side, consumer expectations have been permanently reset by frictionless modern platforms. When everyday life offers on-demand transportation via Uber, instant accommodation via Airbnb, and seamless grocery delivery via Instacart, patients find it intolerable that booking a medical consultation still involves phone trees, paper clipboards, and weeks of delay.

Simultaneously, the incumbent healthcare apparatus—comprising health systems, commercial payers, and self-insured employers who underwrite the majority of private insurance premiums in the United States—has fallen under severe economic siege. Hospitals operate on razor-thin or negative operating margins driven by surging labor expenses, compounded by an unprecedented exodus of clinical staff who suffered severe burnout during the pandemic. Faced with an unsustainable cost structure and an exhausted workforce, healthcare organizations that historically took years to deliberate over enterprise software have been forced into an aggressive search for technological relief.

“I think that gap has really gotten to a breaking point where consumers are saying, I want to vote with my feet and my dollars to get better services wherever I can get them,” Julie Yoo emphasizes. “And on the other side, the incumbent industry is under extreme siege, right? So if you think about large hospital systems, large insurance companies, employers foot the majority of the bill for most insurance fees and premium dollars for a vast majority of Americans. And so all of these organizations are sort of experiencing increased cost structure, right? So labor has gotten much more expensive. Labor has gotten much more constrained.”

- **The Consumer Expectation Gap:** Modern consumers habituated to instant mobile services now actively reject legacy clinical workflows, forcing health systems to compete directly on accessibility and patient experience. - **The Employer Economic Burden:** Because private self-insured employers subsidize the health benefits of most working Americans, escalating premium costs have turned corporate CFOs into aggressive champions of disruptive health tech. - **Post-COVID Labor Shortages:** Widespread clinical burnout triggered a massive exodus of nurses and physicians, creating acute labor shortages that traditional recruitment cannot solve, forcing systems to look toward AI-driven automation.

`10:00` Systemic Pressures, Deductibles, and the Leapfrog Advantage

Healthcare providers and health plans have reached an inflection point where decades of structural bloat have made the traditional business model impossible to maintain. Julie Yoo points out that one of the core drivers making healthcare appear so esoteric is its third-party payer dynamic: because individuals do not directly pay for the vast majority of their care, they historically never felt the friction of each dollar spent, which allowed massive inefficiencies to pile up. Now, both the government and private payers are clamping down hard on runaway costs, forcing incumbent providers to reckon with reality. Compounding this margin squeeze is the extreme scarcity of clinical expertise, which traditionally required at least seven years of intensive post-secondary schooling and specialty training—a threshold software previously could not touch until large language models arrived.

To curb rising expenditures, employers and insurers shifted expenses onto patients through high-deductible health plans. When consumers suddenly had to pay the first $1,000 or more entirely out of pocket before coverage kicked in, they realized how subpar their expensive experiences were. This awakening catalyzed a direct-to-consumer adoption wave for transparent, lower-cost healthcare alternatives. Julie notes that patients can now turn to modern startups offering exceptional clinical services for as little as $10 a month out of pocket. When Sophia Dew asks where AI delivers the most tangible impact across the stack, Julie explains that healthcare’s notorious historical lag in technology spending—ranking among the lowest across all industries relative to revenue—has unexpectedly turned into a structural advantage.

Unlike other sectors that spent tens of billions of dollars over two decades constructing workflow SaaS rails and middleware that now must be painfully ripped out, healthcare stayed largely confined to base-layer electronic health records (EHRs). Consequently, the sector is free from immense sunk cost biases. As Julie explains:

> “We don't have to worry about ripping out 20 years and tens of billions of dollars worth of middleware software. We can actually just go directly to this agentic form of technology.”

```mermaid flowchart LR accTitle: Healthcare Legacy Leapfrog Comparison accDescr: Diagram comparing how typical industries must rip and replace middleware versus healthcare leapfrogging directly to agentic AI.

subgraph OtherIndustries["Typical Enterprise Industries"] O1["EHR / Core ERP Layer"] --> O2["20 Years of Workflow SaaS Middleware"] O2 -->|"Rip & Replace (High Sunk Cost)"| O3["Agentic AI Layer"] end

subgraph HealthcarePath["Healthcare Sector Path"] H1["Basic EHR / Core System"] -->|"Direct Leapfrog (Low Sunk Cost)"| H3["Agentic AI Layer"] end ```

- **Training barrier**: Clinical specialization demands at least 7 years of intensive education, creating an acute labor supply constraint that software historically could not augment until generative LLMs. - **Consumer cost-shifting**: Deductible thresholds of $1,000+ forced employees to pay out of pocket, igniting consumer demand for transparent $10/month subscription healthcare alternatives. - **Underinvestment upside**: Healthcare’s historically low IT spend relative to revenue spared it from the middle-tier workflow SaaS investments that other industries are now forced to replace.

---

`15:00` Organic Clinical Adoption and the Consumer Shift to LLMs

Past waves of health technology adoption were heavily coerced: the federal government had to subsidize doctors through Meaningful Use incentives to adopt electronic health records, and a global pandemic was required to drive telehealth uptake. In contrast, Julie Yoo emphasizes that the generative AI boom is the first completely organic adoption wave in digital health history. Frontline clinicians are actively pulling AI scribes into their daily workflows without top-down mandates simply because it eliminates hours of evening charting. This grassroots, product-led growth (PLG) motion gives doctors unprecedented autonomy to choose tools that reduce administrative burnout.

A parallel revolution is unfolding on the consumer side through platforms like ChatGPT and Claude. Millions of patients now bypass traditional triage to ask language models about symptoms, lab results, and diagnoses. While these models lack formal medical licensure, they offer immediate answers that contrast sharply with the reality of waiting four months for an in-person specialist appointment or paying a $2,000 emergency room bill. Generalist AI interfaces are already beginning to forward-integrate into clinical booking marketplaces, teaching an entire generation of patients that seeking healthcare advice does not have to be an ordeal. As Julie observes:

> “This currently is the only, like the first real organic adoption wave that we've seen in health tech where doctors are just using AI scribes because they freaking love it.”

To bridge the gap between fast consumer chatbot queries and regulated medical care, venture-backed companies are building hybrid, AI-native practices. Julie points to portfolio company Council Health, which provides a 24/7 asynchronous chat environment where AI triage operates alongside licensed medical doctors who hold full authority to diagnose conditions, order lab tests, issue prescriptions, and make specialist referrals. This architecture delivers instant conversational accessibility without sacrificing clinical validity.

---

`17:30` Rare Disease Breakthroughs and the Three Buckets of AI Care

The speed at which generative models parse complex medical information was demonstrated during the "Rare Disease Real Kid Hackathon," organized by Anthropic and AWS with $50,000 in prizes. Sophia Dew highlights how researchers and developers were given the genomic data and clinical records of an actual child suffering from an undiagnosed rare disease (with parental consent). Participants quickly synthesized complex clinical patterns to generate diagnostic leads—a task that typically stalls because rare pediatric conditions suffer from a severe lack of specialist attention and commercial funding.

Julie agrees, noting that while discovery feeds are filled daily with similar breakthroughs, the journey of health AI ultimately spans three distinct buckets:

1. **Access to Intelligence**: Democratizing diagnostic and clinical information so anyone can access top-tier medical knowledge instantly. 2. **Clinical Verification and Execution**: Bridging software insight with licensed testing, in-person procedures, and prescription delivery. 3. **Longitudinal Care**: Transitioning away from episodic queries toward an AI doctor that stays with a patient for life.

Generalist foundation models have largely cracked the first bucket by putting diagnostic knowledge into any smartphone. However, Julie asserts that the most valuable frontier for health tech founders lies in solving the difficult second and third buckets—delivering the verified last-mile medical action and offering continuous care over decades. As Julie Yoo summarizes:

> “I think the first bucket is being solved by a lot of these generalist tools. I think it's really the second and third bucket that therein is the opportunity for health tech builders to really take a crack at that last mile of what do we do about it and how do we stay with you for life until it's resolved, until the next issue comes up, etc.”

```mermaid flowchart TD accTitle: The Three Buckets of AI Healthcare accDescr: Flowchart illustrating the evolution of AI healthcare from access to intelligence, to clinical execution, to longitudinal lifelong care.

B1["Bucket 1: Access to Intelligence<br/>• Solved by generalist LLMs (ChatGPT, Claude)<br/>• Zero-friction medical information & triage"] --> B2["Bucket 2: Verification & Last-Mile Action<br/>• Clinical-grade testing & physical diagnostics<br/>• Licensed MD prescriptions & specialty referrals"]

B2 --> B3["Bucket 3: Longitudinal Lifelong Care<br/>• Shifting from transactional triage to continuous monitoring<br/>• Personalized 'doctor in your pocket' across decades"] ```

- **Rare Disease Real Kid Hackathon**: An Anthropic and AWS event offering $50,000 in prizes that crowdsourced genomic analysis and clinical insights for an undiagnosed child. - **Transactional versus longitudinal**: Contemporary health AI remains largely transactional, but the largest venture opportunity is building lifelong, persistent medical companions that maintain patient context across years of care.

`20:00` The New Opportunity Landscape: Consumers as Payers and AI-Native Full-Stack Companies

Sophia Dew asks Julie Yoo what advice she gives founders seeking the biggest unsolved problems and opportunities in healthcare today. Yoo responds with palpable enthusiasm, remarking that it is “the best time to be building in this space” and admitting she feels jealous of founders getting to build right now. Highlighting recent theses published by Andreessen Horowitz, Yoo points first to the consumer healthcare sector. She jokes that pitching a consumer health business seven years ago would get an entrepreneur practically laughed out of venture offices because no viable business model existed, consumers lacked agency, and payment mechanisms were absent. Today, however, that dynamic has inverted: some of the fastest-growing health tech startups are direct-to-consumer, out-of-pocket cash-pay businesses offering services at radically low costs.

```mermaid flowchart LR accTitle: Healthcare Market Shift accDescr: Comparison of legacy healthcare software design versus the modern consumer-first AI model. subgraph Legacy["Legacy Software Model"] L1["Design for Payers & Providers"] --> L2["Prohibitive Out-of-Pocket Cost"] end subgraph Modern["Modern AI-Driven Model"] M1["Design for Consumers"] --> M2["100x Lower Delivery Cost Structure"] M2 --> M3["Direct Cash-Pay Consumer Products"] end ```

Yoo explains that while paying out of pocket was theoretically possible before, prohibitive pricing made it unrealistic. Artificial intelligence has fundamentally changed that reality by collapsing unit economics. As Yoo observes, “And now because of the magic of AI, the cost structure to deliver an actual medically credible service is like 100 times lower than it used to be.” This cost collapse fuels a flourishing marketplace centered on end-users rather than traditional gatekeepers, a phenomenon detailed in their firm’s essay titled *Consumers Are the New Payer*.

Beyond pure software, Yoo emphasizes businesses that are both "AI-native and AI-proof." While a medical general-intelligence LLM can answer questions that previously required a six-month wait to see a specialist, digital intelligence alone cannot handle the messy physical realities of care. Companies must still draw blood, render physical diagnoses, provide in-person support to patients unable to visit hospitals, and operate compliant, highly regulated legal entities. Echoing the rise of full-stack tech challengers from earlier eras, modern winners resemble classic brick-and-mortar clinics or service retailers on the outside, but run on radically automated, AI-native infrastructure on the inside. This architecture enables superior gross margins and large product surface areas, generating profits that can be continuously reinvested into technological innovation.

- **Robotics and High-Acuity Care**: Yoo identifies genuine product-market fit between modern robotics hardware and the acute, physical demands of high-intensity hospital environments. - **Rethinking Healthcare Reimbursement**: When critics complain that the healthcare system is broken, Yoo counters: “If you actually study the innards of how incentives are aligned based on the payment system that exists in healthcare, it results in exactly what we experience today.” - **Full-Stack Precedents**: Companies such as Devoted Health have shown that founders can essentially build an integrated mini-healthcare system under one roof, enabled by lowering barrier-to-entry costs and cheaper infrastructure.

`25:00` A Ten-Year Prediction: "N of 1" Medicine and an AI Doctor in Every Pocket

Turning to the future, Dew asks Yoo to share a defining prediction for what healthcare will look like a decade from now—the kind of shift someone will tell their future children about. Building on an earlier discussion about individuals analyzing their own genomic data, Yoo predicts a pervasive era of "roll your own healthcare" centered on hyper-personalized, "N of 1" interventions. While these solutions will be uniquely tailored to an individual patient, they will simultaneously benefit from collective intelligence fed back into aggregate health networks.

Yoo points out a fundamental limitation of current medical machine learning: the industry lacks the necessary data to train truly comprehensive clinical-grade models. Existing training sets rely almost entirely on fragmented electronic health records (EHRs) captured during episodic clinic visits, which typically occur only once a year for the average person. As a consequence, the vast continuous narrative of everyday human physiology and patient life is completely unrecorded. Continuous, personalized experiments will lay the foundational data rails required to train the next wave of healthcare algorithms.

Looking forward to the ultimate outcome of these data rails, Yoo declares: “All of us will genuinely have an AI doctor in our pocket for life.” Dew agrees that everyone will have a dedicated, personalized specialist that understands their individual body better than anything else in the world, framing it as the core justification for being a tech optimist. Dew concludes the interview by thanking Yoo for breaking down the future of medicine, followed by standard podcast sign-offs directing listeners to a16z's channels across YouTube, Apple Podcasts, Spotify, X, and Substack.

## Notable quotes > `25:00` All of us will genuinely have an AI doctor in our pocket for life. — Julie Yoo

> `20:00` And now because of the magic of AI, the cost structure to deliver an actual medically credible service is like 100 times lower than it used to be. — Julie Yoo

> `20:00` If you actually study the innards of how incentives are aligned based on the payment system that exists in healthcare, it results in exactly what we experience today. — Julie Yoo

> `05:00` I believe healthcare is the industry that will benefit the most from AI. — Julie Yoo

## Mentioned Julie Yoo, Sophia Dew, Andreessen Horowitz (a16z), Kyruus, Human Genome Project, Stripe, Anthropic, Dario Amodei, Uber, Airbnb, Instacart, ChatGPT, Claude, Council Health, AWS, Rare Disease Real Kid Hackathon, Devoted Health

## Video description a16z General Partner Julie Yoo joins MTS host Sophia Dew to explain why she believes healthcare could benefit more from AI than almost any other industry, and why decades of slow technology adoption may actually give healthcare an advantage in the AI era. Julie traces healthcare’s evolution from paper records and fax machines through electronic health records and telehealth, and explains why AI represents something different: an organic adoption wave driven by tools that doctors and patients actually want to use. Because healthcare never built the same layers of legacy software as other industries, it may now be able to leapfrog directly into agentic AI. They also explore how AI could dramatically lower the cost of care, why consumers are becoming a more important payer, where Julie sees the biggest opportunities for healthcare founders, and a future where everyone has a highly personalized AI doctor in their pocket for life.

[Source](https://mgln.ai/e/1344/afp-848985-injected.calisto.simplecastaudio.com/3f86df7b-51c6-4101-88a2-550dba782de8/episodes/bec89b11-0527-42ce-b965-a903f73f2754/audio/128/default.mp3?aid=rss_feed&awCollectionId=3f86df7b-51c6-4101-88a2-550dba782de8&awEpisodeId=bec89b11-0527-42ce-b965-a903f73f2754&feed=JGE3yC0V)

中文简报

a16z合伙人Yoo：软件欠账反成医疗领域Agentic AI跨越式优势 医疗行业历史上在企业中间件应用上的滞后，使其免于承受巨额的软件替换成本，从而能够直接跨越式迈向Agentic AI，将服务交付成本降低最高达百倍。 在本次讨论中，a16z普通合伙人Julie Yoo深入阐述了为何医疗行业长期以来的软件投入不足反而创造了意想不到的跨越式优势，使该领域得以绕过数十年的SaaS中间件，直接采纳Agentic AI。在临床医生职业倦怠、消费者自付费用上涨以及现代基础模型的共同推动下，医疗领域正通过AI速记员和对话式分诊等工具迎来首次完全自发的采纳浪潮。Yoo概述了三大前沿领域——普惠化智能、经过验证的临床行动以及纵向全程护理，并强调了面向消费者的自费服务以及全栈式AI原生诊所中存在的巨大创业机遇。最终，她预测持续的生理数据采集将推动医疗转向高度个性化的“N=1”护理，让每个人口袋里都能拥有一位相伴终生的AI医生。 医疗行业历来较低的IT支出以及中层企业级SaaS的缺位，使其能够直接从基础电子病历系统跨越到Agentic AI工作流程，避免了成本高昂的推倒重来式软件更迭周期。 (10:00) 与以往自上而下的政策强制令（如“有效使用”或疫情紧急法规）不同，生成式AI浪潮标志着医疗领域首个自发形成的采纳趋势，其驱动力来自临床医生为缩短病历记录时间而积极主动引入AI速记员。 (15:00) AI模型已将提供可靠医疗建议的单体经济成本降低了约100倍，这推动了直接面向消费者的医疗繁荣，患者愿意以透明、低成本的订阅方式自行掏腰包付费。 (20:00) 据预计，在未来十年内，超越每年偶发性门诊的持续个性化数据采集，将为每个人口袋中专属、终身陪伴的AI医生提供动力。 (25:00)

Reading summary

Andreessen Horowitz General Partner Julie Yoo argues that healthcare's historical underinvestment in enterprise software now positions it to leapfrog legacy middleware directly into agentic AI. Driven by severe labor constraints, rising consumer costs, and collapsing software delivery expenses, the industry is transitioning toward organic clinician adoption, direct cash-pay models, and continuous personalized care. The Enterprise Leapfrog and Organic Adoption (10:00) Healthcare historically lagged other enterprise industries in software investment, adopting basic electronic health records (EHRs) only after a multi-billion-dollar federal push subsidized clinicians tens of thousands of dollars each over a five-to-seven-year period. Yoo explains that this delay spares healthcare from having to rip out decades of expensive workflow SaaS rails. Providers can bypass middle-tier software entirely and deploy agentic AI directly on top of EHR foundation layers. Unlike previous government-mandated digitization or pandemic-enforced telehealth adoption, Yoo observes that generative AI is experiencing healthcare's first purely organic adoption wave. Frontline physicians are autonomously pulling AI documentation scribes into their daily practice to escape evening charting burdens, while patients turn to generalist models like ChatGPT and Claude to circumvent long specialist queues and high emergency room bills. Economic Pressures and Consumer Cash-Pay Models (6:48) A convergence of rising provider overhead, pandemic-era clinical burnout, and employer cost-shifting has destabilized traditional healthcare operations. As employers shifted expenses onto workers via high-deductible health plans exceeding $1,000 out of pocket, consumers grew intolerant of phone trees, paper intake, and delayed visits. In response, patients are increasingly paying cash out of pocket for transparent, affordable care, sometimes accessing digital medical subscriptions for as little as $10 per month. Yoo contends that artificial intelligence has reduced the cost structure of delivering medically credible services by roughly 100 times, creating a sustainable foundation for direct-to-consumer healthcare businesses. Where consumer health models were historically dismissed by venture investors due to third-party payer gatekeeping, collapsing unit economics now enable startups to sell directly to patients as the primary payers. AI Frontiers: From Triage to Full-Stack Care (17:30) Yoo divides artificial intelligence healthcare solutions into three progressive stages: democratizing access to intelligence, clinical verification and execution, and longitudinal care. While foundation models have largely addressed diagnostic triage—exemplified by Anthropic and AWS's $50,000 Rare Disease Real Kid Hackathon synthesizing complex pediatric genomic records—Yoo identifies the largest venture opportunities in bridging software insights with physical execution and continuous management. Because pure software cannot draw blood, conduct physical examinations, or maintain regulated healthcare entity licenses, Yoo highlights full-stack care delivery platforms that combine automated back-ends with regulated operations. Startups like Council Health integrate 24/7 asynchronous AI triage with licensed physicians who hold prescribing, testing, and referral authority, pairing automated efficiency with clinical validity. A Ten-Year Outlook on 'N of 1' Medicine (25:00) Looking forward a decade, Yoo forecasts a pervasive shift toward hyper-personalized, 'N of 1' medicine where individuals continuously track and manage their health using dedicated AI companions. Yoo notes that current clinical algorithms are hindered by fragmentary datasets, as conventional models rely on episodic EHR notes logged during visits that occur only once a year for the average individual. By capturing the unrecorded narrative of continuous daily human physiology, patient-driven tracking will supply the foundational data necessary to train clinically rigorous models. Yoo and host Sophia Dew project that this evolution will culminate in every individual carrying a persistent, context-aware AI doctor capable of lifelong longitudinal care. While AI significantly reduces the cost of medical reasoning and access, its long-term efficacy depends on solving the last-mile challenges of physical clinical operations and overcoming the scarcity of continuous, real-world physiological data beyond intermittent clinic visits.

精炼阅读摘要

Andreessen Horowitz（a16z）普通合伙人 Julie Yoo 认为，医疗行业长期在软件技术上的滞后性反而为其带来了跳过数十年企业级中间件、直接跨入AI原生工作流的跨越式优势；随着服务交付成本下降百倍与日常生理数据的沉淀，医疗正转向由消费者驱动、“N of 1”的终身AI伴随照护模式。 历史债务与跨越优势：跳过中间件直达AI (10:00) Julie Yoo 回顾了医疗科技的演进历程：在联邦政府提供数万美元补贴以推行电子病历（EHR）后，美国医疗行业用5至7年勉强告别了纸质档案时代。2010年创立排程平台 Kyruus 时，Yoo 发现患者动辄数月的分诊等待并非源于绝对供给匮乏，而是医院碎片化排程导致医生实际产能利用率远低于100%的路由故障。当时行业缺乏现代云基础设施与标准化API，软件集成如同用凿子与铁钉开凿混凝土墙般艰难。 然而，医疗行业在软件采购上历史投入偏低的现状，在生成式AI时代意外转化为了“后发优势”。Yoo 指出，其他行业必须耗费巨资剥离过去20年积累的SaaS工作流中间件，而医疗行业大部分IT资产仍停留在基础EHR层，没有庞大的沉没成本包袱。这使得医疗生态能够直接绕过复杂的中间件堆栈，跳级采用基于自主智能体（Agentic AI）的全新架构。 制度破冰与自发采纳：技术落地的双向驱动 (5:00) 医疗应用场景的爆发源于政策松绑与现实危机的交织。过去，州级执业许可壁垒和严格的医保报销条例严重限制了远程医疗发展，直至新冠疫情迫使监管部门放宽跨州执业限制并落实远程诊疗同等报销政策，才彻底打破物理诊所边界。与此同时，疫情后临床人员严重倦怠引发离职潮，叠加飙升的人力支出与微薄的医院运营利润率，迫使传统医疗机构必须通过自动化技术填补劳动力缺口。 在需求端，习惯了即时移动消费体验的公众，对传统诊所繁复的电话树和纸质登记板产生了强烈抵触。更关键的是，Yoo 强调生成式AI构成了数字医疗史上首个完全“自发性”的采用浪潮：与过去依赖政府行政命令或外部危机不同，临床医生为了免除晚间繁琐的书写病历负担，通过产品驱动增长（PLG）模式主动引入AI记录员（AI scribes）；数以百万计的患者也开始绕开漫长的专科挂号，直接借助 ChatGPT 和 Claude 进行日常化验单解读与初步分诊。 医疗AI的三大阶段与全栈交付壁垒 (17:30) 结合 Anthropic 与 AWS 举办的罕见病黑客松（以5万美元奖金众包真实患儿的复杂基因组分析）等进展，Yoo 将医疗AI的演进划分为三个阶段：第一阶段是“智能获取”（Access to Intelligence），即利用通用大模型实现医学知识与基础分诊的零门槛普及；第二阶段是“临床验证与执行”（Clinical Verification and Execution），负责衔接处方开具、物理检验与合规医疗操作；第三阶段则是“连续性照护”（Longitudinal Care），推动医疗从单次事务性咨询跃升为伴随患者终身的连续照护。 尽管通用大模型已基本攻破第一阶段，但 Yoo 认为健康科技最具商业价值的护城河集中在后两阶段。由于单纯的软件无法替代抽血、体检或重症监护等线下履约环节，最有前景的赢家将是“AI原生且防AI替代”（AI-native and AI-proof）的全栈实体。例如 Council Health 结合AI分诊与持牌医师构建闭环，或如 Devoted Health 般将技术底层与合规医疗网络融为一体，在外部呈现为传统医疗实体，内部则借助AI自动化维持极高的毛利率与服务响应效率。 消费者支付浪潮与“N=1”终身AI医生 (20:00) 高免赔额健康计划（个人自付门槛普遍达1000美元以上）的普及，迫使承担自费压力的消费者主动寻找透明且平价的就医替代方案。Yoo 表示，AI将提供合规医疗服务的边际成本削减了约100倍，直接推动了以每月10美元等极低门槛切入的直接自费（Cash-pay）消费级医疗模式崛起，打破了以往健康科技只能向保险公司或雇主兜售方案的单一商业逻辑。 展望未来十年，Yoo 预测医疗将迎来“N of 1”的超个性化精准时代。当前医疗机器学习的核心瓶颈在于高质量训练数据严重匮乏——绝大多数算法仅依赖患者每年一次偶发门诊所记录的断裂EHR数据，完全错失了日常连续的生理变化叙事。随着日常个性化监测轨道的铺设，群体的聚合智能将与个体数据紧密交织，最终使每个人都能在口袋中拥有一位终身伴随、深度理解自身独特体质的AI医生。 尽管 Julie Yoo 对AI重构医疗生态前景表达了高度乐观，但该愿景仍面临现实约束：当前医疗AI应用仍多局限于事务性浅层交互，缺乏物理闭环与长期随访机制；现有算法严重受制于碎片化的门诊历史数据，如何在严苛监管下构建包含血液检测、实体诊疗在内的可信履约体系，并将日常生活中的生理信号转化为受认可的临床决策，仍是AI医疗由辅助工具迈向全栈医生的核心挑战。

Detailed notes — zh

## Background 本期节目由 a16z 通用合伙人 Julie Yoo 与 MTS 节目主持人 Sophia Dew 于 2026 年 9 月 6 日在播客及音频平台联合发布。节目简介指出，医疗行业因常年技术采用缓慢而缺乏繁琐的遗留软件包袱，如今反倒有望在人工智能时代实现跨越式发展，迈向由 AI 驱动降本增效、消费者成为新型支付方，以及人人拥有口袋专属终身 AI 医生的未来。 ## TL;DR Julie Yoo 结合自身近二十年创立 Kyruus 与主导医疗投资的历程，深入拆解了医疗行业从传真机时代、电子健康档案（EHR）普及到跳过传统 SaaS 中间层、直接跨入 Agentic AI 的独特演化路径。她指出，传统第三方支付制度的失效以及高免赔额健康险的推行，正倒逼普通消费者成为关注性价比的新支付主体，催生出自费医疗与 7×24 小时人机协同诊疗等创新模式。面对未来，Julie 强调医疗科技创业的核心在于跑通诊断后的临床落地闭环并构建全栈防 AI 淘汰壁垒，最终依托 N=1 超个性化健康数据轨道，实现人人拥有终身专属口袋 AI 医生的愿景。 ## Atmosphere & style 对谈整体充满技术乐观主义与前沿创投的亢奋感，Julie Yoo 以兼具深厚软件工程底色与医疗老兵沉淀的视角侃侃而谈，言语间洋溢着“嫉妒当今创业者”的由衷热情。两位对话者配合默契，从昔日用凿子敲混凝土般的艰涩回忆，层层推进至当下自发性 AI 采用潮与罕见病黑客松的激动人心案例。整场交流既有对传统体制弊病的锐利剖析，又具备描摹未来生命健康蓝图的磅礴张力，听来令人振奋且充满启发。 ## Bulletin - `13:16` 医疗行业因缺乏沉重的传统 SaaS 中间层与沉没成本负担，有望跳过传统企业软件阶段，直接跨越式进入以 Agentic AI 为原生的工作流。 - `20:00` 大语言模型将具备医学专业可信度的服务边际成本降低了近百倍，直接推动自费医疗模式爆发，使消费者成为核心的新买单方。 - `13:16` AI 医疗转录听写（AI scribes）成为医疗科技史上首次完全由医生自发且由产品驱动增长（PLG）带动的有机采用浪潮。 - `25:00` 未来真正的终极医学大模型无法仅靠稀疏离散的传统 EHR 数据训练，需依赖 N=1 的超个性化工具持续采集高密度全周期健康数据。 ## Key takeaways - `00:00` Kyruus 创办的核心假设在于化解“患者就诊矛盾”：普通患者预约挂号需等待数月，而医院内部医生的工时利用率却严重不足。 - `05:00` 早期医疗科技处于缺乏 API 和基础设施的“石器时代”，打通底层老旧系统极度困难，但新冠疫情迫使医疗生态重构了跨州执业与远程支付等核心假设。 - `07:30` 日常消费软件（如 Uber、Airbnb）拔高了公众体验预期，使得传统医疗的低效与滞后逼近临界点，促使消费者开始主动用脚和钱包投票。 - `10:00` 第三方支付机制割裂了消费者的财务痛感，而高免赔额健康计划的推行迫使患者自付大额费用，激发出庞大的自费高性价比医疗需求。 - `13:16` 医疗行业科技采纳缓慢的传统反倒成为优势，使机构没有复杂的 SaaS 迁移包袱，能够直接跃迁至智能体（Agentic AI）生态。 - `15:58` 面对昂贵的急诊室与漫长的专科医生排期，ChatGPT 和 Claude 已不可逆地成为大众低摩擦获取即时健康建议的前哨入口。 - `15:58` Council Health 等平台通过 AI 对话原生界面与持证全科医生的结合，打通了 7×24 小时开立处方、诊断和转诊的医疗合规闭环。 - `17:48` 底层通用大模型解决了获取医学智能的门槛，而医疗创业者的核心阵地在于搞定临床落地的“最后一公里”以及伴随全生命周期的健康管护。 - `20:00` 单纯依赖通用大模型容易被底层迭代吞噬，当前最顶尖的医疗初创企业倾向于构建“全栈挑战者”模式，将 AI 原生流程深度嵌合进实体服务与资质闭环中。 - `25:00` 现存电子健康档案（EHR）仅记录每年零散的离散切片，严重缺失患者在真实生活中的全周期动态体征与健康叙事。 - `25:00` 超个性化端侧应用将充当新型数据轨道，推动未来每个人都能拥有一个装在口袋里、全天候陪伴且极度懂你的专属终身 AI 医生。 ## Detailed walkthrough ### `00:00` 医疗数字化的缘起与“患者就诊矛盾” 医疗是全球成本最高的行业之一，然而患者的时间与优质医疗资源却始终极度匮乏。a16z 通用合伙人（General Partner）Julie Yoo 在做客 MTS 节目与主持人 Sophia Dew 对话时指出，医疗很可能是全行业中从人工智能（AI）获益最深的一个领域。长久以来该行业技术采纳缓慢的痛点，如今反倒成了优势——它让医疗行业得以绕过复杂的传统企业软件堆栈，直接跳跃到以 AI 为原生的工作流中。 ```mermaid flowchart TD accTitle: 医疗行业数字化跃迁历程 accDescr: 展示医疗从早期纸质办公、政府激励推行EHR，到Kyruus解决供给矛盾，最终跨入AI原生阶段的演进过程。 subgraph Phase1["数字化早期 (约20年前)"] A["纸质文件夹与传真机办公"] --> B["联邦政府资金补贴政策<br/>(数万美元/医生)"] B --> C["普及电子健康记录 (EHR)<br/>建立核心记录系统"] end subgraph Phase2["互联网与平台期 (2010年前后)"] D["人类基因组计划与计算生物学爆发"] C -.-> E["创立 Kyruus"] D -.-> E E --> F["解决患者就诊矛盾<br/>(挂号排队数月 vs 医生工时利用不足)"] end subgraph Phase3["现代AI跃迁期"] F --> G["跳过繁冗遗留软件层"] G --> H["直接进入 AI 原生工作流与 Agentic AI"] end ``` Julie Yoo 回顾了自己近 18 年的医疗科技从业生涯。她最初是一名训练有素的软件工程师，职业生涯前七年深耕更广泛的科技生态，写代码并经历了早期电子商务、金融科技（Fintech）和制造业的兴起。约二十年前，正值美国联邦政府着手推动医疗数字化之时，她转型进入该领域。当时的医疗机构甚至还在使用纸质文件夹和档案柜，毫无数字化可言。随后联邦政府推出了极具魄力的补贴政策，为每位落地电子健康记录（EHR）系统的医生提供数万美元补贴。在随后的五到七年里，几乎 100% 的美国医生都在埋头推行这类核心记录系统（System of Record）。 Julie Yoo 先后涉足了计算生物学与基因组学领域（受“人类基因组计划”爆发带动），随后于 2010 年联合创办了 Kyruus 并担任首席产品官（CPO）长达九年。创办 Kyruus 的核心假设源于一个令人震惊的行业悖论：全美各大主要城市中，普通患者预约一次初级保健医生的等待时间长达数周甚至数月；然而，当她与大型医院和医师集团的管理层交流时，对方却透露其医生的时间利用率远低于 100%。Kyruus 的目标正是化解这一“患者就诊矛盾（Patient Access Paradox）”，打造覆盖整个医疗系统的统一号源调度平台，精准引导患者完成预约匹配。 - 联邦政府通过直接补贴方案，驱动全美医生耗费 5 至 7 年全面部署电子健康记录（EHR）。 - 创办 Kyruus 的核心痛点：患者挂号需苦等数月，而医院内部的医生利用率却长期严重不饱和。 - Julie Yoo 谈及当年转型初衷时表示：“Now it's time for healthcare to have its shot in the light.”（现在轮到医疗行业迎来高光时刻了。） ### `05:00` 医疗科技的“石器时代”与后疫情时代的破局 在创立 Kyruus 的年代，医疗科技领域几乎是一片荒漠，Julie Yoo 将其形容为医疗科技的“石器时代”——那是一个没有 Stripe、缺乏基础 API 层的时代。当时愿意涉足医疗风投的机构寥寥无几，她开玩笑说当时全行业可能就五个投资人，而她把这五个人的钱都拿了一遍。为了将软件接入医疗老旧的底层设施，她付出的努力就像“拿着凿子和钉子硬敲混凝土墙”。不仅资金与时间成本极高，人才招募也极为艰难，毕竟工程师往往倾向于去 Google，而不是去翻新 1970 年代的老旧系统。然而坚守十余年终于迎来回报，公司在去年成功售出，她戏称这是一场“历时 15 年的一夜成名（15 year overnight success）”。 卸任公司职务后，Julie Yoo 加入了 a16z 负责医疗健康投资团队。她观察到过去三到五年间，几股巨大的驱动力彻底打破了原有的僵局。首先是新冠疫情（COVID-19），它倒逼整个医疗生态重构第一性原理，撕毁了原先刻意限制供给的多项假设： 1. **执业许可松绑**：以往美国实行严格的州级执照制度，医生极少拥有跨州行医资格；但远程医疗（Telehealth）带来了全国规模的临床服务需求，促使监管机构全面反思并放宽执业壁垒。 2. **支付模式变革**：过去很多远程医疗服务若无既往面对面问诊的前提则不予报销，这一规则在疫情中被迅速打破。 最核心的转折点则是 AI 的涌现。Julie Yoo 大胆断言，医疗将是全行业中从 AI 获益最大的领域。过去数年间，AI 在医疗领域的采用与爆发速度充分印证了该转折点的到来。 - 基础设施断层：早期医疗科技处于“Pre-API”时代，必须完全依靠硬编码硬凿底层接口。 - 人才与资本迁移：从原先极少基金关注、难与大厂竞争技术人才，演变为目前科技界头部力量（如 Dario Amodei 领导的 Anthropic）公开重注医疗健康、立志攻克癌症等复杂顽疾。 - 谈及疫情对固有医疗体制的冲击，Julie Yoo 指出：“And it really tore apart a lot of the previously held assumptions about how healthcare needs to work in terms of everything from how are physicians credentialed to practice medicine.”（它真正打破了此前关于医疗该如何运作的诸多既定假设，涵盖了从医生行医资格认证到其他一切环节。） ### `07:30` 医疗复兴：消费者觉醒与承保机构的成本围城 针对主持人 Sophia Dew 关于“医疗行业为何迎来当下这场复兴”的提问，Julie Yoo 分析指出，这是多重社会与经济因素交织汇聚的必然结果。医疗是少数能够 100% 影响每个个体生命周期的行业，但绝大多数人在医疗中的体验往往令人沮丧。 与此同时，生活其他领域已被现代化科技彻底重塑：Uber 改变了出行，Airbnb 颠覆了旅宿，Instacart 革新了食品杂货采购。大众对现代服务业的期待水准已被拔高到全新维度，而医疗行业的体验却常年停滞不前。这种反差已经逼近临界点，促使普通消费者开始“用脚和钱包投票”，主动探寻体验更佳的医疗服务。 另一方面，传统医疗巨头正身陷重度围城（Under Extreme Siege）。无论是大型医院系统、保险巨头，还是为绝大部分美国雇员承担保险费及保费支出的雇主群体，都在面临空前的成本结构承压。医疗劳动力的薪资成本水涨船高，加之疫情带来的极端压力造成了医护人才的大规模流失，行业面临极度严峻的人力短缺。多重重压之下，整个行业被迫主动敞开大门，迫切寻求能实现降本增效的技术解药。 - 消费者心理预期差：日常便捷软件（Uber、Airbnb、Instacart）的体验跃升，使得传统医疗的低效与滞后愈发不可容忍。 - 支付方与机构承压：为绝大多数美国人买单的雇主及保险公司面临飞涨的保费开支。 - 劳动力流失危机：疫情压力导致医护人员大批离职，劳动力短缺进一步推高了行业的人力成本。 - 谈及全行业对技术改造的迫切渴望，Julie Yoo 强调：“So I think that gap has really gotten to a breaking point where consumers are saying, I want to vote with my feet and my dollars to get better services wherever I can get them.”（我认为这种差距已经逼近临界点，消费者正在明确表示，他们要用脚和钱包投票，无论哪里能提供更好的服务就去哪里。） ### `10:00` 医疗成本转嫁与传统支付体系的崩塌 Julie Yoo 深入分析了当前医疗系统所面临的体制性重压。长期以来，医疗行业之所以显得极其晦涩且效率低下，根源在于奇特的“第三方支付系统”（third-party payer system）：绝大多数医疗账单并非由患者自己掏腰包支付，导致消费者对每笔支出的痛感钝化，进而催生出远超自由市场机制的巨量系统冗余与膨胀。然而，随着医护人员因繁重负担纷纷转行，政府与商业保险公司对不可持续的成本膨胀展开严厉施压。面对支付模式的结构性危机，传统医疗提供方与保险机构不得不承认现行路径已走到尽头——正如 Julie Yoo 所断言：“Our business model like no longer works, like the math doesn't math anymore.” 作为回应，系统将成本转嫁给消费者。过去雇主承担大部分保费，而如今高免赔额健康计划（High-Deductible Health Plans）迫使个人承担切身经济责任（skin in the game）。当必须自付 1,000 美元免赔额后，消费者猛然惊醒：为什么花如此大价钱，买到的却是糟糕透顶的服务体验？这种切肤之痛促成了一场由下而上的自费服务浪潮。市面上迅速涌现出一批极具性价比的创新医疗平台，用户每月仅需自付 10 美元左右，就能获得比传统医疗机构好上千倍的体验。加之传统医疗需要至少七年的严苛医学规培才能产生专家级人才，这种高门槛在过去让软件难以模拟其专长，而今天大语言模型（LLMs）的突破终于填补了供需两端的鸿沟。 - **医疗成本膨胀根源**：第三方支付机制掩盖了真实服务定价，割裂了消费者体验与财务痛感。 - **免赔额机制的冲击**：高免赔额（如自付前 1,000 美元）让患者转变为关注性价比的真正消费者。 - **自费医疗服务生态**：以每月约 10 美元为代表的新兴自费模式，服务体验大幅碾压传统昂贵体系。 --- ### `13:16` 医疗行业的“跨越式技术跃迁”与有机自发采用 当主持人 Sophia Dew 问及医疗如何比其他行业从 AI 中获益更多时，Julie Yoo 提出了一个逆向视角的“跨越式发展”（leapfrog dynamic）假说。历史上，医疗行业在科技支出与营收占比上堪称全行业垫底，对技术采纳近乎排斥。然而，这在 AI 时代反而成了一种罕见的优势。金融、企业服务等现代行业在过去二十年构建了庞大繁重的 SaaS 工作流和中间件层，耗费了数百亿美元，并训练了整整一代员工使用这套界面；如今面对智能体（agentic AI），他们必须承担巨大的沉没成本并忍痛推倒重来。相比之下，医疗体系只有底层的电子病历/ERP 架构，中间层基本是空白的，因此完全没有“沉没成本偏见”（sunk cost bias），可以直接一步跨入 Agentic AI 时代。 更关键的是，这一轮技术应用展现出前所未有的自发性。Julie Yoo 回顾了以往被动的技术推广史：早年政府需要用补贴去求医生使用电子系统，后来必须靠全球疫情才能迫使大众尝试远程医疗。但如今正在发生的是医疗科技领域“第一次真正的有机采用浪潮”。医生们之所以疯狂采用 AI 医疗转录听写（AI scribes），纯粹是因为它能解决实际痛点，正如 Julie Yoo 所强调的：“This currently is the only like the first real organic adoption wave that we've seen in health tech where doctors are just using AI scribes because they freaking love it.” 这一现象推动了个体医生和一线临床人员依靠产品驱动增长（PLG）模式自主拥抱技术。 ```mermaid flowchart TD accTitle: Healthcare Technology Leapfrog accDescr: 医疗行业跳过传统 SaaS 中间层直接跃迁至 Agentic AI 的对比流程图 subgraph OtherIndustries["其他典型行业（传统技术沉重）"] A["底层 ERP / 基础设施"] --> B["庞大的 SaaS 工作流与中间件"] B --> C["高额沉没成本与技术替换阵痛"] C --> D["被迫重构为 Agentic AI"] end subgraph HealthcareIndustry["医疗行业（跨越式路径）"] E["底层 EHR / ERP 记录系统"] E -.->|"无沉没成本负担<br/>直接跳过中间层"| F["原生 Agentic AI 与智能体生态"] end ``` --- ### `15:58` 消费级 AI 医疗应用与全天候人机协作实践 在消费者端，ChatGPT 和 Claude 已经成为大众获取健康建议的首选入口。即便通用大模型的医学建议尚未获得正规行医执照与严格审批，用户依然蜂拥而至。原因在于传统就医摩擦力过大：患者需要权衡是等待长达四个月去预约专科医生、支付 2,000 美元去急诊室，还是直接掏出手机在低阻力渠道获取即时反馈。这种体验代差不可逆地重塑了消费者的心智预期。 针对这一趋势，Julie Yoo 旗下的 a16z 正在重注新一代消费级医疗企业，其中典型代表便是其投资组合中的 Council Health。该平台提供了一种 7×24 小时的异步 Chat 模式，但其核心创新在于：与用户聊天的不仅仅是一个冰冷的 AI 机器人，后台无缝结合了拥有真实行医执照的注册内科医生（MD）。持证医生在对话流中可以即时开立处方、安排转诊并做出权威临床诊断，巧妙实现了 AI 原生效率与实体医疗合规救治的有机整合。 - **替代权衡逻辑**：急诊室耗费 2,000 美元或专科号排期 4 个月，使得大模型聊天工具成为低摩擦替代方案。 - **Council Health 模式**：AI 原生对话界面结合持证执业医生，实现 7×24 小时开方、转诊与诊断闭环。 - **消费级体验演进**：大模型正逐步前向整合进现实医疗服务网络与医生预约交易平台。 --- ### `17:48` 罕见病黑客松与医疗 AI 创业者的“最后一公里” 谈及 AI 的临床诊断突破，Sophia Dew 分享了 Anthropic 与 AWS 联合举办的“罕见病真实患儿黑客松”（Rare Disease Real Kid Hackathon）。该赛事设立了 50,000 美元奖金，在家长知情同意下公开了一位患儿的全基因组与临床数据，让全球参赛团队利用 AI 生成前沿诊疗洞见。鉴于罕见病专科医生资源极度匮乏且研发缺乏商业利润支撑，AI 在此类边缘重症领域展现出不可估量的生命拯救价值。 Julie Yoo 对此深有共鸣，指出社交媒体上每天都有海量类似案例涌现。她将医疗 AI 的发展划分为三个递进维度： 1. **获取智能的门槛（Accessing Intelligence）**：由底层通用大模型解决，消除了认知阻碍与信息不对等； 2. **临床落地闭环（What to do about it）**：在获得推断后，如何进行临床级验证检查、开具处方药物、完成线下手术介入与复杂流程； 3. **长期纵向跟踪（Longitudinal Care）**：超越现存单次、事务性的对话，提供伴随个人一生的专属口袋医生。 Julie Yoo 最终点明了医疗科技初创企业的核心机会：“I think the first bucket is being solved by a lot of these generalist tools. I think it's really the second and third bucket that therein is the opportunity for health tech builders to really take a crack at that last mile of what do we do about it and how do we stay with you for life until it's resolved, until the next issue comes up, etc.” 解决从洞察到治疗的闭环与全生命周期的健康管护，正是未来医疗创业的制胜阵地。 ### `20:00` 医疗创业的新机遇：消费者成为新买单方与“全栈防 AI 淘汰”模式 当主持人 Sophia Dew 问及从创业者转型为投资人后、站在宏观视角会对寻找方向的医疗创业者提出什么建议时，Julie Yoo 难掩兴奋地感叹，现在的时机实在是太好了：“How much time do we have? I mean, there's the opportunity. It's like the best time to be building in this space. I'm so jealous of all my founders who get to be building right now.”（我们有多少时间细聊？机会实在太多了。现在绝对是在这个领域创业的最佳时代，我甚至无比嫉妒所有正在这个风口上开拓的创始人。）她首先指出的核心突破口就是 C 端消费者市场。在七年前，如果向投资人推销面向消费者的医疗健康项目，创业者大概率会被直接轰出会议室，因为当时既没有可行的商业模式，消费者缺乏掌控自身健康的主动权，更不存在顺畅的支付渠道。而如今，HealthTech 领域增长最快的公司大多采用自费模式（Cash-pay）——消费者直接掏腰包以极其低廉的价格购买服务。 Julie Yoo 强调，AI 带来的核心杠杆在于彻底颠覆了成本结构，使得交付具备医学专业可信度的服务成本比过去下降了近百倍。以往即使消费者愿意自费，高昂的价格也令人望而却步；而现在的全新产品从第一天起就是为消费者本人设计，不再像过去那样不得不去讨好保险公司和医生体系。a16z 为此专门撰写了题为《消费者是全新的支付方》（*Consumers Are the New Payer*）的研究文章来阐释这一范式转移。与此相辅相成的另一大趋势是构建“AI 原生（AI-native）且具备防 AI 淘汰能力（AI-proof）”的企业。仅仅拥有一个能回答全科医学问题的通用大模型是不够的，患者面临的现实问题依然严峻：无法前往医院时如何获得线下面诊？如何抽血化验以确诊疾病？如何合规地开展现行法律监管下的医疗行为？因此，当前最出色的公司往往采用类似上一代科技创业中的“全栈挑战者（Full-stack Challenger）”模式：从外表看，它们可能只是一家普通的医疗服务机构或零售商，但底层却完全由 AI 驱动，从而具备了传统手段无法企及的业务覆盖面与极高的利润率。 在支付体系层面，Julie Yoo 驳斥了外界对于美国医疗系统“陷入混乱”的流行抱怨，指出整个系统并非出差错，而是按照既定设计在运转：“If you actually study the innards of how incentives are aligned based on the payment system that exists in healthcare, it results in exactly what we experience today.”（如果你真正去研究医疗健康领域基于支付体系所形成的激励机制内核，就会发现当前的糟糕体验恰恰是这种机制运作下的必然结果。）正因如此，许多雄心勃勃的初创公司（例如 Devoted Health 等）正尝试从零开始重构支付体系，打造集支付与医疗服务于一体的微型医疗系统。基础设施的普及与 AI 的杠杆效应，极大降低了这类宏大构想的入局门槛。 ```mermaid flowchart TD accTitle: AI Native Healthcare Value Creation accDescr: 展示 AI 原生医疗企业如何打通从低成本交付到全栈闭环与重构支付的链路 subgraph CostDisruption["成本结构颠覆"] A["生成式 AI / 医疗模型"] --> B["专业级服务成本降低近 100 倍"] B --> C["消费者自费 (Cash-Pay)"] end subgraph FullStack["全栈防 AI 淘汰壁垒"] C --> D["外部：实体机构 / 穿戴与硬件机器人"] D --> E["内部：AI 原生流程与高毛利架构"] end subgraph SystemReinvention["系统重构"] E --> F["重塑激励机制与支付底层 (如 Devoted Health)"] end ``` - **自费模式爆发**：AI 让医学级服务的边际交付成本降低了约 100 倍，消除了自费医疗的价格门槛，诞生了一批绕过保险机构、直接面向消费者的医疗健康公司。 - **全栈 AI 模式的优势**：单纯依靠软件界面的大模型容易被底层技术迭代吞噬，而结合了线下护理、化验诊断、合规资质以及硬件机器人的“AI 原生 + 全栈服务”构筑了深厚的竞争壁垒。 - **硬件与高急性期护理**：在对护理要求极高的重症与高急性期（High-acuity）医疗场景中，机器人及硬件解决方案正在快速找到产品市场契合点（PMF）。 --- ### `25:00` 十年预言：N=1 超个性化医疗与口袋里的终身 AI 医生 当被问及对未来十年的医疗健康格局有何大胆预测时，Julie Yoo 认为个性化医疗将迈入彻底由患者自己主导的时代。针对 Sophia 提到的自行分析基因组数据的例子，Julie 指出未来将涌现出大量“自主构建医疗（Roll-your-own healthcare）”的应用，即所谓的“N=1”超个性化解决方案——它完全围绕个人特质定制，同时又能从全体用户沉淀汇聚的群体智慧网络中受益。 谈及医疗大模型的训练瓶颈，Julie Yoo 提出了一个深刻的行业洞察：现阶段全行业甚至根本没有足够的数据来训练出真正的医学级模型。现有的医疗数据高度依赖电子健康档案（EHR），但这些记录极其稀疏零散。绝大多数普通人一年顶多去一次传统诊所，这就导致患者生命周期中长段的情境、日常体征和生活体验在现存数据集中是彻底缺失的。随着前述各种 N=1 级个性化健康工具的普及，它们将充当全新的数据轨道，持续采集高密度的真实健康信息，从而反哺并训练出下一代医学大模型。 Julie Yoo 坚信这一演进最终会带来颠覆性的就医体验：“All of us will genuinely have an AI doctor in our pocket for life.”（我们每个人都将真真正正拥有一个装在口袋里、陪伴终身的 AI 医生。）Sophia 赞同地补充道，届时每一个人都会拥有一位比世界上任何人都更了解自己的专属专家。这正是技术乐观主义者对未来充满憧憬的最坚实根基，科技正在切实推动这一未来的到来。 - **EHR 数据的根本缺陷**：传统电子健康记录仅记录每年极少数就诊节点的离散切片，严重缺失患者在日常状态下的全周期健康叙事弧线。 - **N=1 数据轨道**：超个性化端侧应用正在承担生成新型实时连续健康数据的重任，为构建未来真正的终极医学模型奠定基础。 - **口袋里的终身专家**：未来的 AI 医生将摆脱传统医疗体系挂号等候数月、诊断碎片化的困境，成为全天候陪伴、深度理解用户生理特征的专属全科与专科医生。 ## Notable quotes > `25:00` All of us will genuinely have an AI doctor in our pocket for life. — Julie Yoo > `20:00` If you actually study the innards of how incentives are aligned based on the payment system that exists in healthcare, it results in exactly what we experience today. — Julie Yoo > `00:00` Now it's time for healthcare to have its shot in the light. — Julie Yoo ## Mentioned Julie Yoo, Sophia Dew, a16z, Kyruus, Council Health, Anthropic, Dario Amodei, AWS, Devoted Health, Google, Stripe, Uber, Airbnb, Instacart, ChatGPT, Claude ## Video description a16z General Partner Julie Yoo joins MTS host Sophia Dew to explain why she believes healthcare could benefit more from AI than almost any other industry, and why decades of slow technology adoption may actually give healthcare an advantage in the AI era. Julie traces healthcare’s evolution from paper records and fax machines through electronic health records and telehealth, and explains why AI represents something different: an organic adoption wave driven by tools that doctors and patients actually want to use. Because healthcare never built the same layers of legacy software as other industries, it may now be able to leapfrog directly into agentic AI. They also explore how AI could dramatically lower the cost of care, why consumers are becoming a more important payer, where Julie sees the biggest opportunities for healthcare founders, and a future where everyone has a highly personalized AI doctor in their pocket for life. [Source](https://mgln.ai/e/1344/afp-848985-injected.calisto.simplecastaudio.com/3f86df7b-51c6-4101-88a2-550dba782de8/episodes/bec89b11-0527-42ce-b965-a903f73f2754/audio/128/default.mp3?aid=rss_feed&awCollectionId=3f86df7b-51c6-4101-88a2-550dba782de8&awEpisodeId=bec89b11-0527-42ce-b965-a903f73f2754&feed=JGE3yC0V)

Source quotes

[25:00] Julie Yoo: All of us will genuinely have an AI doctor in our pocket for life. [20:00] Julie Yoo: If you actually study the innards of how incentives are aligned based on the payment system that exists in healthcare, it results in exactly what we experience today. [0:00] Julie Yoo: Now it's time for healthcare to have its shot in the light.

Source quotes

[25:00] Julie Yoo: All of us will genuinely have an AI doctor in our pocket for life. [20:00] Julie Yoo: And now because of the magic of AI, the cost structure to deliver an actual medically credible service is like 100 times lower than it used to be. [20:00] Julie Yoo: If you actually study the innards of how incentives are aligned based on the payment system that exists in healthcare, it results in exactly what we experience today. [5:00] Julie Yoo: I believe healthcare is the industry that will benefit the most from AI.

Original transcript

[0:00] Healthcare is one of the most expensive industries in the world, yet time and medical expertise remain scarce. Julie Yu thinks AI could change that. The A16Z general partner joins Sophia Du on MTS to explain why healthcare may benefit more from AI than almost any other industry, and why its slow adoption of technology could now be an advantage, allowing it to leapfrog legacy software and to move directly to AI-native workflows. from consumer healthcare and AI native care delivery to robotics and new payment models, and why the future could include an AI doctor in your pocket for life. Julie, welcome to MTS. Thanks so much for having me. I'm so excited to chat with you and I'm so excited to get into this conversation because often in some of the conversations around AI, people don't totally understand how it affects them. And I think healthcare is one of the most obvious places where you actually get to see that possible. But before we jump into all of the details there, I would love to just start a little bit with your personal story. Why is healthcare care industry now for probably, gosh, 18 years at this point. And I was too early. You know, like back then, I was actually a technologist by training. So out of school was a software engineer in the broader tech ecosystem, wrote code for the first seven years of my career, did a lot in like the early era of like e-commerce and fintech and manufacturing, etc. And then made a career change into health care right when the federal government was starting to talk about how do we digitize health care? Because I'm sure I've talked to anyone about their experience in health care. One of the first things they'll say is I'm being paper to death, We still have a long way to go. But imagine 20 years ago when the majority of the industry was actually still working on paper, like doctors offices still had those paper folders in their cabinets and nothing was digitized. And so the federal government decided to implement this massive program, which is kind of a miraculous program when you think about it. Basically, what they said was we will pay doctors tens of thousands of dollars to implement electronic health records. And that was the way by which our industry sort of came into the modern era was for over the course of about five to seven years, I would say. Literally 100% of doctors in the U.S. were heads down implementing these kind of system of record, electronic health record platforms. And that's how we kind of came into digitization. So right around then was when I made a career change thinking, okay, I've now seen what great looks like with respect to enterprise technology being applied to all these other industries, e-commerce, retail, manufacturing, etc. Now it's time for healthcare to have its shot in the light. about. And I initially started out in kind of the genomics era, which was one of the first areas where you saw just an explosion of data. We had the Human Genome Project. We now understood the fundamental code of humanity. And there was a whole industry that exploded around kind of computational biology. How does genomics tie to ultimate disease and things of that sort? So did a couple of really cool companies in that area. And then ultimately around 2010 was when I started Kyrus, which was my company, really with the thesis of some of what you described. So we as issues when we try to get in touch with the healthcare system. The average wait time for a primary care doctor's appointment is weeks, if not months in most major cities across the US still today. And when me and my co-founder started to talk to executives who were running these large hospitals and physician groups in the industry, what we didn't realize was they were telling us, yes, people are waiting out the door. But if you actually look at the way that our physicians are being utilized, we're still way under 100%, which was shocking, right? We had this paradox of And yet when you looked at utilization rates of physicians' time, it was way under 100%. And so that was the thesis behind our company, Kairos, was how do you sort of solve this patient access paradox by creating a single system of appointment inventory across the healthcare system and the ability to route patients to the right doctor and the right appointment in such a way that you could actually effectively utilize that capacity much more efficiently. And so I did that for an amazing nine years as co-founder and CPO. sort of like the stone age of health tech. So this gets to sort of my current job today. Back then was imagine tech sort of like pre-Stripe, pre all the sort of like API layer of how you build companies and stand up companies. That was effectively the situation in health tech back then. Very little venture capital also going into our space back then. And a lot of the conventional wisdom at the time was healthcare is just so hard, right? It's a highly regulated industry. Yeah, I think that's a good point. [5:00] So back then, I joked that there was like five guys who like invested in our space and I raised from like all five of them. But, you know, it was pre-APIs. I liken the work that I did literally with my own two hands as taking a chisel and a nail against this concrete wall of integrating into the infrastructure of health care. And so it took a lot of time, a lot of capital. It was really hard to recruit talent back then because it just was an unsexy industry, right? If you had a choice of going to Google versus trying to build these 1970s systems, like why would you choose to come into our space? So we did it the hard way, but great outcome. We actually just sold the company last year. Congratulations. Thank you. Yeah. So huge. Yeah. And a decade paid off. It was like a 15 year overnight success. But in that time, so many things occurred to really mature the industry in so many ways that now I would argue this is our moment in health tech. And I've now been an investor after I stepped on from my company. I came to the dark side and I now run the health care investing team at Indreason. And what's occurred over the last The one, obviously we had COVID, which I think was a huge reason why people were forced to recognize the need to create a digital modality for engaging with the healthcare system. And it really tore apart a lot of the previously held assumptions about how healthcare needs to work in terms of everything from how are physicians credentialed to practice medicine. Historically, it's been a very sort of local system where we sort of deliberately constrain supply in healthcare in weird ways. So you have to get licensed at the state level. not the case that doctors typically would have a license to practice medicine outside of their state. But with telehealth, all of a sudden you had this national scale demand for clinical services. And so lots of states and federal agencies had to kind of rethink from first principles, how do we relax some of these constraints to enable people to get access to care when you can no longer go in person? That was one. Another was kind of the way that we pay for health care. Telehealth previously was in many cases was not paid for unless you had previously had an in-person lot of those roles. So there was a lot of tailwinds in the industry during that period that sort of led to acceleration of the ability to build tech companies in this space. And then obviously AI. I mean, I think I've declared loudly that I believe healthcare is the industry that will benefit the most from AI for many reasons that I'm sure we'll talk about. But the level of accelerated growth and scale and adoption that we've now seen in the last, let's call it three to five years post AI, I think is proving that point and demonstrating that the moment is here Definitely. It's so fascinating getting to hear your story because you understand how technology has affected the healthcare industry for over multiple decades. And you've gotten to see when you're way too early, as you say, I like how you mentioned your 15-year overnight success. it wasn't as sexy of an industry. It was harder to get talent. It was harder to get capital. But it seems to me that that's no longer the case. Even if what you're seeing with Dario, where he's saying, okay, we're doubling down in health, we're going to cure cancer. I talk to startups all of the time, especially in AI. And one of the common themes that we're seeing is that they're going towards the hard problems, the ones that are really bureaucratic, billions, billions of lives. And so I'm so excited to see that. And I would love to just get your opinion. What shifted? What was the big change over the past few years that gave us this renaissance, this healthcare industry renaissance that we're in? I think it was a confluence of factors. So, I mean, exactly to your point, it's always been the one industry that literally impacts everyone. Every single life has some interaction with the healthcare system at some point. And typically it's a very negative experience, unfortunately. And so I think one where, especially as all of the experiences in the rest of our life, right, if you think about how Uber has changed transportation, how Airbnb has changed travel, how Instacart has changed the way we think about food, et cetera, et cetera, our standard and our bar for what we expect from any service industry has gone way up while healthcare has kind of stayed stagnant. So I think that gap has really gotten to a breaking point where consumers are saying, I want to vote with my feet and my dollars to get better services wherever I can get them. the incumbent industry is under extreme siege, right? So if you think about large hospital systems, large insurance companies, employers put the majority of the bill for most insurance fees and premium dollars for a vast majority of Americans. And so all of these organizations are sort of experiencing increased cost structure, right? So labor has gotten much more expensive. Labor has gotten much more constrained. We did unfortunately have a big sort of outflux of talent from the healthcare industry because of COVID, you know, because it was such a stressful situation. [10:00] an arduous time that a lot of people said, I'm throwing the towel in and I'm going to go do something else with my medical training. So at the same time, the government has also really come down on the cost, right? And said, we have just unsustainable amounts of bloat right now. And this is a little bit of a rat hole, but one of the reasons that healthcare appears to be so esoteric is that we have this weird third-party payer system, right? Like the vast majority of your healthcare is not paid for by you. And so you don't necessarily feel the pain with every dollar and therefore there tends to be a lot more bloat than if you would have with like a free market dynamic. And so, you know, the government's coming down on that. Payers in general are coming down on that. So you have all this cost pressure on the incumbent system that has driven, you know, traditional providers and payers to say, our business model like no longer works, like the math doesn't math anymore. And we can't continue to sustain this kind of payment modality. back where, you know, historically, because healthcare is such an esoteric, sort of scarce training oriented industry, where you literally have to go to school for what is it, you know, at least seven years and train and to even become a specialist, you know, service provider in our industry. So the bar to, you know, being able to create technology that could replicate any, anything that's close to that level of expertise has been really difficult until we now have this, with LLMs. And so I think that, you know, those I would say are the major compulsive factors that have now resulted in a recognition by both the supply side and the demand side that things need to change. And I mean, the last thing I'll say is that one of the ways by which the system had reacted to this kind of cost bloat in the past was to say, okay, if you have, let's say, insurance covered by your employer, you know, historically, the employer covered the majority of the costs, ago was, okay, let's actually do some cost shifting. Let's actually give consumers a little bit of skin in the game and say, okay, we're going to have deductibles. And so up to $1,000, your insurance is not going to kick in. So anything under that you're going to have to pay for. And then that actually is what caused consumers to realize, okay, gosh, I have to pay how much for what kind of crappy services? And people really start to feel it directly. And now you have this proliferation of services to kind of fill that void, right? We have amazing companies that literally $10 per month out of pocket to get incredible services that when compared to what you could get for that same amount of dollars with the traditional healthcare system is just a thousand times better. And so that's what really, I think, has caused the adoption wave that we've seen by just individuals. Oh, interesting. So what I'm hearing is there's some systemic shifts around how healthcare is just paid for, and that's really been changing over the last few years. And then, of course, what exists now with technology and AI. And one of the things that you argued is that healthcare may benefit more from AI than arguably any other industry. And so I want to just unpack a little bit more about where exactly you can see AI's benefits when it comes to healthcare. One of the areas that we were just talking about yesterday is in discovery of diseases and drug discovery, consumer until maybe it goes through a bunch of clinical trials. I know there's a lot of different angles. If we could just briefly touch upon what The whole staff, where is AI actually playing? How will I as a person, as a consumer, as a customer, How will I fuel this benefit? Yeah, absolutely. I'll talk about it in a couple of different ways. One is the industry itself has been almost like antithetical to technology adoption historically. We literally have been like the lowest in terms of dollars spent on tech relative to revenue, like the lowest of the lowest in terms of industries. And so, you know, one thing that's changed now is kind of this like what I call a leapfrog dynamic, where when you think about every other industry, like workflow SaaS rails, right? Like software service companies. And they had this whole sort of like middleware layer on top of infrastructure that they spent tens of billions of dollars on, trained a whole generation of workers on. And then when AI came along, they said, oh my God, we have to rip and replace this whole previous paradigm of technology and replace it with agentic AI and retrain our whole generation of workers on that modality. Healthcare for better, for worse, did not spend that amount of money, right? So we only had really like the ERP layer And so in some ways, we have less of a sunk cost bias as an industry to say, okay, this new technology paradigm has come along, and we can just leap right to it. We don't have to worry about ripping out 20 years and tens of billions of dollars worth of middleware software. We can actually just go directly to this agentic form of technology. And that's actually been a huge boon to many of the companies that sell into it. [15:00] the space. So I wrote a blog post about a year ago that talked about the fact that historical eras of technology adoption in healthcare were always very unnatural. Like what I talked about earlier, you had to pay doctors to adopt technology. You had to have a pandemic to force people to use telehealth. This is currently is the only like the first real organic adoption wave that we've seen in health tech where doctors are just using AI scribes because they freaking And you're seeing a lot more of this almost like PLG led motion where individual physicians and frontline workers actually have agency and being able to use these tools in a much different way. So that's one part that's happening more on kind of the enterprise side. On the consumer side, you know, thank you, Chachi, PT and Claude. Like everybody is going to LLMs for health related questions. Right. And they're pretty damn good. Right. And they've gotten a lot better since the early days when they first came out. advice and recommendations. And they're starting to forward integrate into actual marketplaces where you can get advice from ChatGPT and then go book an appointment with a real doctor. And so I think that's just taught a whole generation of humans. There's number one, just a much more low friction way to get medical advice. Even if the medical advice might not be regulated and licensed and fully validated in the way that you would get from doctors, the trade-off is I either have to wait four months to get a real doctor's appointment, or go to an emergency room and pay $2,000 or I can just open up my phone. And so I think that tradeoff, again, that delta makes it incredibly easy for people to just go to the lowest friction place. And so I think people like consumers are now educated that there's a better way. And now we have a whole wave of new companies, many of which we've invested in, that are taking a specialized approach to how we leverage AI to create consumer-grade experiences, called Council Health in Our Portfolio, which is literally an AI native doctor practice that is a 24 by 7 asynchronous chat modality with real doctors. Like imagine chatting with your chatbot, but having an actual licensed MD in the chat who can fully prescribe, who can fully refer, who can fully diagnose. And so you kind of get the best of both worlds of, you know, AI native care with real doctors who can actually provide clinical utility. And so you're seeing even just a few years ago. One of the things that just wasn't possible a few years ago that I wanted to bring up, there was this hackathon by Anthropic AI and AWS Cloud. And it was called the Rare Disease Real Kid Hackathon. And they had $50,000 in prizes. And the way that they did it is they actually had a child, they uploaded the kid's genome and the clinical data, and they shared it with the community, of course, with parental consent. And people were able to actually... Understand and generate insights. And I saw this and I was I was I was. amazed. This is so cool. It's really hard to find specialists, especially for rare diseases. Those are children with rare diseases. There's so little resources actually going into this. The money doesn't actually make sense. And so seeing this, I was like, wow, this is amazing. This is one really clear example of how AI can actually play out, can actually save human lives. saw it. Yeah, no, absolutely. I mean, I think every day, I mean, one of the most exciting things is that like literally every day on my X feed, I will see a hundred examples of something like that. And I think to me, there's sort of three parts. One is, you know, how do you lower the barrier to access intelligence, which again, that barrier used to be a real, you know, kind of impediment to people getting access to healthcare. Now that thing, that's really gone away. And then I think now the question moves towards what do you do about it, right? Once you actually understand a or what diagnosis you might have, you still need to, number one, get that verified. You might need actual testing or kind of clinical grade testing to get it fully diagnosed. You might need to be either prescribed a therapeutic or a procedure or get some sort of in-person care. And then there's the longitudinal aspect of it as well. A lot of the way that people are using, which is great, I mean, medical AI today is highly transactional. Imagine if you can have a doctor for life that's I think the first bucket is being solved by a lot of these generalist tools. I think it's really the second and third bucket that therein is the opportunity for health tech builders to really take a crack at that last mile of what do we do about it and how do we stay with you for life until it's resolved, until the next issue comes up, etc. [20:00] side to now the investor side, you one, have to make a bit of predictions about where things are going to go. And two, you get to see the opportunity landscape almost at a macro level. And so if someone is coming to you for maybe advice or there's a founder who is asking, okay, Julie, where do I go? What are the big opportunities here? What are the big problems that still need solving? What advice would you share? Yeah. How much time do we have? I mean, there's the opportunity. It's like the best time to be building in this space. I'm so jealous of all my founders who get to be building right now. I think probably a couple that I would highlight that we've written about recently so folks can kind of check out our theses on this. One is definitely the consumer angle. I joked a few months ago that if you had come to any investor, call it, you know, seven years ago and pitched a consumer health business, you would have been kicked out of the room because there was no viable business model. It was not clear that consumers understood and had agency to, you know, really take control of their health, etc. And, you know, the payment models were not there. So now that all that's changed and some of the fastest growing companies that we see in health tech are, you know, cash pay. You just pay out of pocket for a service at a disruptively low price. I think that's actually the biggest lever that's changed Even if you could pay for things out of pocket historically, the price was prohibitive. And now because of the magic of AI, the cost structure to deliver an actual medically credible service is like 100 times lower than it used to be. And so you're seeing a whole proliferation of an entirely new marketplace of products that were designed for the consumer, not like historically where you had to design for the insurance company and the doctor. exciting way. So we wrote a piece called Consumers Are the New Payer, basically, to sort of recognize that trend. So that's one big area. Another post that I had just last week, actually, was one where we made the argument that the best companies right now are those that are AI native and AI proof. And it kind of gets to what I was saying earlier, where yes, you can have an amazing, you know, medical, sort of general medical intelligence, LLM, that answers a bunch of wait for six months to see a specialist to get. But what about the rest of it? Right. So how do you actually deliver in-person services to people who can't travel to the hospital? How do you actually do, you know, how do you take a blood draw and actually diagnose like what you actually have? How do you, you know, do all the things that today, at least in our country, are not like legal to do, you know, lest you have some sort of regulated entity. So we have a lot of companies that on the outside, and this kind of mirrors a lot of what we saw in the first era of more broadly, where you had kind of these challenger businesses that were full stack in nature. So like from the outside, they might just look like a retailer or just look like some kind of service provider. But on the inside, they were obviously highly AI native and therefore had a surface area that was not achievable, you know, through historical means and a cost structure that was disruptive such that they can continue to invest in innovation because they had high margins. We're seeing a ton of companies that are embracing kind of the best of both worlds in that way. combination is a lot of where the excitement is. Hardware companies are included in that. So I think robots are, I think there's product market fit between what robotics companies can do and what is necessary in really high-acuity healthcare settings. And then the final thing I'll say is insurance, the payment rails are, you know, the payment One of the reasons that people look at healthcare and they say, oh, it's so screwed up. Why is it so screwed up? I'm like, it's actually designing exactly as planned. If you actually study the innards of how incentives are aligned based on the payment system that exists in healthcare, it results in exactly what we experience today. And so there's a lot of really ambitious companies that are saying, what if we blew up that current healthcare payment modality and built a new one from scratch? you could do is kind of build almost like a mini healthcare system in one company. But there's companies like Devoted Health and others who are doing this. And I think now it's possible because, again, the cost to entering into building a company in our space has gone down because all this infrastructure now exists. You can do so much more with AI. And the appetite from the system, both in terms of consumer customers as well as the traditional ecosystem, is there in a way that's Last question for you. What prediction do you have that will be true in a decade from now? I would love to one day, you know, tell my kids or future kids, yeah, back in my day, healthcare was X and now. Yeah, yeah. What do you think? I mean, I think so getting to your example about the. [25:00] roll your own genome analysis thing. I do think there will be a lot of roll your own healthcare, where you will have N of one solutions that are only super personalized to you, but still get the benefit of the broader set of intelligence that everybody's contributing back to. I mean, one of the interesting things about healthcare is that in order to train a model in healthcare, I would argue that we actually don't even have the datasets necessary to truly train a medical-grade electronic health record data, which is really sporadic, right? Like how many times you actually go to see a traditional doctor in your life, it's like maybe at best once a year. So there's a whole narrative arc of your experience as a patient that is absent from pretty much every data set that exists today. So I think a lot of companies will start doing these kind of N of one experiments that generate the new data rails that then drive the next generation of models that are built in All of us will genuinely have an AI doctor in our pocket for life. Yeah, we'll all have a personalized specialist that knows us better than anything in the world. That is so exciting to see. And this is one of the reasons I am such a big tech optimist and so excited for the future and how technology is impacting the future. And you touched upon directly some very concrete reasons. But Julie, really wanted to thank you for joining us today on MTS. Absolutely. Such a great conversation. Thanks for having me. Thank you so much for joining. listening to this episode of the A16Z podcast. If you like this episode, be sure to like, comment, subscribe, leave us a rating or review and share it with your friends and family. For more episodes, go to YouTube, Apple Podcasts and Spotify. Follow us on X at A16Z and subscribe to our Substack at a16z.substack.com. Thanks again for listening, and I'll see you in the next episode. or any of its affiliates. Information is from sources deemed reliable on the date of publication, but A16Z does not guarantee its accuracy. Bye. Thank you. you