Summary

Molly O’Shea says AI has crossed from experimentation into measurable healthcare impact after a decade of watching it produce few additional drugs or therapeutics. “Everything has changed in the last two years,” led by biological and chemical models, laboratory robotics, and simulations that could raise success rates in the most expensive, failure-prone product-development industry.
Devoted Health is O’Shea’s strongest proof that AI compounds a hard-won operating moat rather than replacing one. The company spent a decade securing state licenses, provider networks, distribution and primary-care infrastructure; over the past year it tripled in size, halved operating expenses and significantly increased EBITDA. Her call is unusually categorical: it is “probably the best example of AI in healthcare in the world right now.”
Healthcare’s first AI dividend should be augmentation, with much of the value accruing invisibly to patients. Administrative work consumes 25-30% of healthcare costs, while O’Shea says roughly 30% of radiology images are read incorrectly; AI can reduce both burdens without eliminating doctors. Patients may never credit AI, but they should feel it when clinicians stop typing with their backs turned and miss fewer findings.
O’Shea frames the US-China life-sciences race as an existential industrial risk, not a routine competitive cycle. She says roughly 50% of pharmaceutical companies’ external research dollars now go to China, whereas this was not happening five years ago, while China may already match or exceed US university patent output. Her response combines AI investment with stronger university research and “a complete reorganization of our FDA and processes.”
AI has raised the required quality of founders faster than it has relaxed the need for engineers. O’Shea wants technically fluent, product-oriented entrepreneurs who can change course, recruit humbly and understand distribution; the best create conviction “within five minutes.” The career lesson she still carries is blunt: “Your bar for entrepreneurs is not high enough.”
Only about 10% of today’s AI valuation inflation is justified in O’Shea’s view. A handful of companies can address enormous markets spanning software and services, but many niche businesses valued at $10 billion or $20 billion may ultimately be worth only $1 billion or $2 billion and acquired on EBITDA. “Only 10% of the companies in the world are worth it, and 90% aren’t.”
Oak is pairing unusually flexible capital with company formation in markets where AI can execute, not merely analyze. Its funds grew from $500 million to $2 billion, checks range from $1 million to $100 million, and 20-40% of a fund can go to early-stage investments. Augur received $100 million at launch to build supply-chain agents, while Halluminate’s simulated financial environments went “from zero to a hundred in six months.”

Deep dive

1. Healthcare AI has finally reached O’Shea’s threshold for real impact

Oak’s healthcare and fintech practices cover roughly half the economy, giving horizontal AI companies access to unusually concentrated buyer sets. Healthcare is “the smallest of the big markets”: O’Shea counts only about 100 consequential providers and 10 major payers, making procurement knowledge and relationships a genuine distribution asset.
O’Shea had studied healthcare AI for a decade without seeing it generate more development candidates or therapeutics. “In the last two years, everything has changed,” prompting Oak’s investment in Chai Discovery and plans to add partners in San Francisco, where she sees much of the relevant AI work concentrating.
The enabling shift is larger than language models: biological and chemical models, physical AI and laboratory robotics can redesign discovery through clinical trials. Drug development is exceptionally expensive and failure-prone, so even a higher probability of success or shorter development cycle becomes “simply a game changer.”
In a later roadmap discussion, O’Shea describes an antibody effort as already about 100 times more productive than a lab without it, followed by peptides and then small molecules. Peptide progress is significant but, as she carefully notes, “there’s nothing they’re ready to show clients yet.” The transcript does not clearly identify this effort as Chai Discovery, so the attribution should not be made categorical.

2. China turns drug discovery into an industrial-policy emergency

O’Shea rejects the comfortable view that China merely tweaks generic drugs. She says Chinese universities and research units now produce as many or perhaps more patents than their US counterparts, while American university research—the foundation of domestic discoveries—faces reduced NIH investment and research funding.
Her warning is explicitly forward-looking: “This is not a problem today. I’m just afraid it will become a bigger problem in the future.” Weakening university research and international participation could damage the pipeline before the loss becomes visible in marketed medicines.
The capital migration is already stark in her account: roughly 50% of external pharmaceutical research spending now goes to China, something she says was not happening five years ago. COVID also exposed dependence on China for approximately 90% of antibiotic ingredients.
The proposed counterweight is not isolation but accelerated domestic capability: computational drug design, wet-lab validation, animal studies, AI infrastructure and regulatory reform. Models must be tested against physical results, but O’Shea believes they are approaching the outputs of development methods used for 40 years.

3. Devoted shows why AI rewards companies that already own the workflow

Devoted’s apparent AI acceleration rests on a decade of expensive groundwork: Medicare Advantage licenses state by state, local provider networks, broker and direct distribution, a proprietary payer platform, and Devoted Medical Group as a virtual primary-care layer. AI enters a system that already assumes both insurance risk and responsibility for care.
Over the past year, O’Shea says Devoted tripled the size of the company, halved operating expenses and materially increased EBITDA. The result is a business she considers “almost impossible to compete with,” particularly as it prepares to enter the commercial market.
Her end state combines an AI interface, a physician behind it and an insurer handling payment and process seamlessly. Asked whether that makes Devoted the leading healthcare example, she does not hedge: “I’m not usually one to exaggerate, but it’s true.”
Siddharth asks whether entrenched hospitals, customer ties and clinical institutions could delay disruption. O’Shea says hospitals are not disappearing; the nearer opportunity is stripping administrative work from physicians, using electronic-health-record data as the substrate and letting clinicians focus on patients.

4. Augmentation should improve care before patients realize AI did it

Administrative activity represents 25-30% of healthcare costs, according to O’Shea. AI may be the first major technology to reduce that burden rather than add to it, reversing the electronic-record experience in which doctors often face a laptop instead of the person they are treating.
On diagnostics, O’Shea says approximately 30% of radiology images contain something radiologists miss. Her answer is not to replace radiologists—“we probably need more of them”—but to put AI behind them so fewer consequential findings escape review.
Siddharth argues that people will still want people treating them, while computer vision and robotics could shorten procedures, improve what surgeons can see and potentially enable expert-guided remote surgery.
Siddharth points to rural cancer care and surgery, where outcomes trail leading hospitals and a local surgeon may have performed a procedure twice rather than 1,000 times. Remote control and robotics could distribute elite procedural capability; he describes today’s da Vinci system as “so brute force.”

5. Founder quality remains the underwriting constant

O’Shea starts with trust: is this the best person to build the company, and do they combine “intelligence and intent”? Because technology and competition move so quickly, she believes a much larger share of funded founders must now perform near the top 0.1%, even if not literally within it.
She disputes the first-wave claim that product generalists would eliminate the need for engineers. Chai’s founders bridge computer science, chemistry and biology; Devoted’s leadership combines computer science with decades in healthcare. Technical depth must coexist with go-to-market understanding and the humility to recruit missing expertise.
Repeat founders provide Oak’s most powerful signal. Todd and Ed Park are examples from athenahealth; Todd later co-founded Castlight, while Oak led a Series A in Devoted. Brad Smith followed his palliative-care model with CareBridge and Main Street. O’Shea’s praise for Smith is specific: “You’re building something that no one else is doing” amid a market crowded with copycats.
An early mentor, retail investor Jerry Gallagher, corrected O’Shea’s own process: she had emphasized theories, ideas and numbers too heavily. His line—“Your bar for entrepreneurs is not high enough”—became permanent guidance, especially when rapid growth from zero to 10 can seduce investors before they evaluate the people.

6. Flexible capital matters, but valuation discipline matters more

Oak began with a $500 million fund and expanded its last two funds to $2 billion as company capital requirements and opportunity sets grew. It can invest $1 million while committing up to $100 million from day one, or enter at seed and continue across subsequent rounds.
Early-stage investments represent 20-40% of a fund—roughly $400 million to $800 million at current scale. Oak can invest up to 10% of a fund in common stock and has also made PIPE investments, preserving a full-lifecycle mandate rather than forcing every thesis into a single financing stage.
O’Shea sees a “barbell-like world” of seed specialists and late-stage or crossover capital. To remain useful to ambitious founders, Oak must write large checks, yet early entry is increasingly important when fast growth can lift valuations beyond a price the firm considers fair.
Her discipline is exit-based: many AI niches lack the TAM or strategic buyer required to support $10 billion-$20 billion marks. Some may become highly profitable, but buyers of narrow administrative software may still value it on EBITDA; O’Shea says Oak can sometimes sell assets at a 30% strategic multiplier when there is a strategic buyer.

7. Execution agents and M&A broaden the investable AI surface

Halluminate builds reinforcement-learning environments in which agents perform the full work of financial analysts: constructing models, forming recommendations, applying scorecards and reviewing related legal work. One simulation reproduced an entire real-estate brokerage so customers could test alternative business models; O’Shea says the company went “from zero to a hundred in six months.”
Augur emerged after Oak spent substantial time developing ideas with Dave Clark, who had spent 22 years building Amazon’s supply chain and logistics. Its layer does more than identify disruptions: agents execute responses across the supply chain in real time. With about eight large Fortune 500 clients entering its second year, Oak funded the company with $100 million at launch.
AI also expands healthcare’s buyer universe beyond McKesson, Cardinal and the major payers. O’Shea points to Microsoft’s acquisition of Nuance and Oracle’s purchase of Cerner as evidence that technology groups may seek deeper exposure to healthcare, pharmaceuticals, life sciences and provider software.
O’Shea estimates M&A integrations still fail more often than they succeed. CareBridge’s expansion inside Elevance is a positive case, but the general rule is strategic clarity: Siddharth notes that a product or data acquisition can work without retaining the team, while continued innovation usually cannot. O’Shea’s integration test is: “Do you care about creating the best enterprise, or are you in a culture of political intrigue?”

The Future of Frontier Model Architectures with Walter Goodwin, Founder & CEO of Fractile Preview: All About Agents — Dots Arrive Before OpenAI Is Ready