🚀 Mission View
A sharper perspective on this week's issue that matters at the intersection of health and AI.
Andrej Karpathy, one of the original architects of modern AI and former director of AI at Tesla and OpenAI, recently offered a dose of realism in a sea of hype. He argues we’re not in the year of AI agents, but the decade of them. The tools we have today are still interns. They get the easy things right but stumble on nuance, novelty, and judgment. Getting from “good demo” to “trustworthy tool” requires what he calls the march of nines — the slow, grinding path from 90 percent reliability to 99.9 percent, where every extra nine of accuracy takes just as much work as the one before it (i.e. the amount of time and effort it takes to go from 90% to 99% is the same as it will take to go from 99% to 99.9%). It’s a reminder that technological leaps are rarely overnight; they’re built one percentage point at a time.
That slow, steady path sounds familiar to anyone in healthcare. Yet, Menlo Ventures reports that 22 percent of healthcare organizations now use domain-specific AI, a seven-fold jump over last year and far ahead of the broader economy, where fewer than one in ten firms have adopted AI at all. Health systems lead the way at 27 percent adoption, using AI for charting, billing, documentation, and early clinical applications that make the work faster and more consistent. That’s unusual for healthcare, a sector that typically lags years behind in adopting new technologies. Think about paper intake forms every time you go to the doctor's office. Need I say more?
Still, a methodical, thoughtful approach is necessary when it comes to AI and health. And for purpose-driven health organizations the goal may not be building the biggest models or operating the flashiest labs, but they’re positioned to do something more meaningful: show how augmented intelligence can be applied responsibly, ethically, and effectively inside complex systems that serve people. For some, yes, the next decade may be about chasing artificial general intelligence; but for others, it will be about building dependable systems that strengthen the humans already doing the work — the kind of progress that mission-driven organizations are uniquely equipped to lead.
🛜 Field Signals
A quick hit on this week’s key policy shifts and industry trends.
The American Medical Association announced a new center to guide responsible adoption of AI and digital tools in clinical practice. The initiative will focus on physician education, evidence-based integration, and policy development to ensure AI improves care without adding burden.
Google Research, along with partners at the University of California, Santa Cruz Genomics Institute and other federal and academic researchers, announced in a new paper, “DeepSomatic: Accurate somatic small variant discovery for multiple sequencing technologies” a tool that leverages machine learning to identify genetic variants in tumor cells more accurately than current methods.
A new study of Microsoft 365 Copilot use across 90 U.K. health organizations found time savings of 43 minutes per staffer per day — the equivalent of millions of hours annually if scaled across the National Health Service. More via The Telegraph.
Axios reports that a genetic testing startup has launched a new AI genomics research arm focused on predicting the likelihood that IVF embryos could develop cancers, Alzheimer’s, and other chronic diseases later in life. As the article notes, “assigning embryos risk scores is expensive and controversial.”
Anthropic launches Claude for Life Sciences. The new platform connects Claude to scientific tools like Benchling and BioRender, adding AI “skills” for lab protocols, data analysis, and regulatory drafting. Researchers can now use Claude to summarize literature, generate hypotheses, draft study documents, and even analyze genomic data — all in one workspace.
STAT: European oncology experts roll out guidance for use of large language models in clinical care.
Google launched Skills, a learning platform featuring 3,000 AI and technical courses, with gamified features and employment pathways through company partnerships.
Public figures across tech and politics have signed a Future of Life Institute letter demanding governments prohibit superintelligence development until it's proven controllable and the public approves its creation.
OpenEvidence — dubbed the “ChatGPT for doctors” — just raised $200M at a $6B valuation, doubling its worth in three months. The platform, used by 40% of U.S. doctors, delivers evidence-based medical answers sourced from peer-reviewed journals like NEJM and JAMA. Its rise underscores a key shift: investors are betting on specialized, high-trust AI built for professional decision-making, not general chatbots.
🛠️ Practical Edge
Actionable tips and tools to help leaders strengthen capacity and apply AI in their work.
HBR: A Systematic Approach to Experimenting with Gen AI
Also from HBR, this article: How Gen AI Can Create More Time for Leadership
OpenAI unveiled Atlas, a new AI-powered browser that blends agent functionality, persistent memory, and seamless app integrations — a direct signal that the next wave of competition will be about who owns the user’s daily workflow.
Microsoft rolled out new Windows 11 updates, adding a voice-activated “Hey Copilot” command and Vision features that let Copilot see your screen and take actions. Anthropic followed with a Microsoft 365 integration, giving Claude direct access to SharePoint, Outlook, and Teams — plus enterprise search across company data. So what? AI is moving from the browser to the desktop, making it harder not to use. For health organizations, that means governance and training will need to keep pace with wherever staff are already working.
🌅 On the Horizon
A quick look at the developments and events expected to shape the weeks ahead.
Oct. 30: CHAI Webinar: Validating AI for Chronic Heart Failure, Real-World Lessons. Register here.
Nov. 18-20: AI in Healthcare & Pharma Summit 2025 (RE•WORK) in Boston. Register here.
Dec. 8-9: Fortune Brainstorm AI San Francisco. Apply to attend here.
Feb. 11-13: The Complexities of AI in Health Care by American Health Law Association in Las Vegas and Virtual. Register here.
⌚️ Closing Time
A parting thought on what health leaders need to be focused on.
Pew’s new global survey on AI sentiment captures the mood: nervousness still outweighs optimism. Among 28,000 adults across 25 countries, half of respondents in the U.S., Italy, Australia, and Greece said they feel uneasy about rising AI use. The EU emerged as the most trusted regulator (53% confidence), while younger adults remain far more optimistic than older generations. For those of us steeped in the health and tech world, it’s a reminder that enthusiasm isn’t universal. As AI moves deeper into care delivery, trust will matter as much as performance. Reminder: innovation may move fast, but trust doesn’t.
Till next time,
