Americans are already using AI for healthcare in huge numbers. According to a recent OpenAI report, more than 40 million people ask ChatGPT health-related questions every day—and one in four of the platform's 800 million users submits a healthcare prompt weekly. Two announcements last week suggest this is just the beginning.

OpenAI launched ChatGPT Health, which lets users load their medical data (from Apple Health or their electronic health record) so the AI can give personalized guidance on symptoms, test results, and preventive care. Meanwhile, Utah approved a pilot program allowing AI (not a physician) to refill certain low-risk medications. Author Robert Wachter (a physician and healthcare AI expert) argues that while both seem modest on their own, together they signal something bigger: the beginning of AI moving from a supporting role to actually delivering care.

Shubham Saboo (AI Product Manager at Google) argues that the traditional PM workflow—gathering customer needs, writing specs, handing them to engineers—is compressing fast:

"When agents can take a well-formed problem and produce working code, the PM's job shifts. You're no longer translating for engineers. You're forming intent clearly enough that agents can act on it directly. The spec is becoming the product.

...The cycle times that used to define product development, from quarterly planning, monthly sprints, to weekly releases are compressing into something closer to continuous deployment of ideas."

The new PM skillset centers on three areas:

Saboo's six guidelines for curating context are particularly valuable, as context directly drives the quality of the output. And if you've never worked the way he describes, he's got practical tips to get started.

On January 12, Anthropic released Claude Cowork as a "research preview." Cowork is essentially a friendlier version of Claude Code aimed at non-technical users. It lets you point Claude at a folder on your computer, queue up multiple tasks, and let it work in the background—reading documents, building dashboards, even filling out web forms via a Chrome extension.

In Allie K. Miller's lightning demo, you can get a feel for Cowork's capabilities. She had it analyze 320 podcast transcripts and build an interactive dashboard while she worked on other tasks in parallel threads. This showcases Cowork's multitasking and task-queuing capabilities. That said, it's early days—there are glitches, plugin support isn't fully available yet, and it's currently only available with the Max plan. Still, Cowork is definitely worth watching as it matures, as it represents a more user-friendly entry into the world of agentic AI for business users.

There is no shortage of AI-related predictions for AI in 2026, but I particularly like Patricia Reiners' concise list of UX predictions for 2026:

Her whole post is worth a read, and she links to a recent podcast episode where she goes deeper into these shifts.

Thanks for reading, and see you next Wednesday with more curated AI/UX news and insights. 👋

All the best, Heidi