Knit Health Launchs With $11.6M Seed to Build AI Based on Real-World Clinical Decision-Making

Healthcare AI companies have largely focused on training models on medical literature, clinical notes, and text-based data. But Knit Health is taking a different approach: teaching AI systems how healthcare actually operates inside hospitals and clinics.

The University of California, Berkeley spinout has emerged from stealth with $11.6 million in seed funding co-led by Uncork Capital and Frist Cressey Ventures, with pre-seed backing from Moxxie Ventures and participation from Coalition Operators. The company says the capital will support development and deployment of its Large Clinical Behavior Model (LCBM), a system designed to learn from how clinicians make decisions across real healthcare environments.

Rather than functioning like a traditional medical chatbot or documentation assistant, Knit Health is building what it describes as “collective clinical intelligence” — AI trained on patterns embedded in patient routing, referrals, scheduling decisions, discharge timing, and care coordination workflows across hospitals.

Moving Beyond Text-Based Healthcare AI

Most generative AI systems in healthcare today are fundamentally language models. They excel at summarizing records, generating notes, or answering questions based on published medical knowledge.

Knit argues that many of the most important operational decisions in healthcare are not explicitly written down. Instead, they emerge from years of clinician experience navigating real-world constraints such as specialist availability, referral bottlenecks, hospital capacity, and patient complexity.

The company’s LCBM is trained using Truveta electronic medical record data spanning more than 130 million patients across 30 U.S. health systems. Knit says it applies techniques including deep reinforcement learning, causal inference, and behavioral cloning to model how care decisions unfold in practice.

This differs significantly from conventional healthcare AI systems that rely primarily on static datasets or published research. Instead of predicting the next word in a sentence, Knit is attempting to predict operational care decisions inside health systems.

According to the company, the system can adapt to the specific operational dynamics of individual hospitals, including referral patterns, staffing limitations, and workflow structures.

Building an Infrastructure Layer for Hospitals

Knit Health is positioning its platform as a foundational intelligence layer for healthcare operations rather than a standalone application.

The company says its models are initially being deployed for triage, patient flow optimization, discharge prediction, referral management, and quality improvement initiatives. Over time, the broader goal appears to be embedding AI into the operational infrastructure beneath nearly every clinical workflow.

This aligns with a broader shift occurring across healthcare AI, where companies are increasingly targeting operational inefficiencies rather than solely focusing on diagnostics or conversational assistants.

Healthcare systems continue to struggle with issues such as delayed referrals, overcrowded specialty care pipelines, inefficient scheduling, and fragmented coordination between departments. These operational problems often directly affect patient outcomes despite advances in clinical knowledge and treatment availability.

Knit’s strategy suggests that future healthcare AI systems may become less focused on replacing physicians and more focused on orchestrating the complex systems surrounding patient care.

Truveta’s Expanding Role in Healthcare AI

Knit’s partnership with Truveta also reflects the growing importance of large-scale real-world clinical datasets in healthcare AI development.

Truveta has built one of the largest collections of de-identified clinical data in the United States, representing more than 130 million patients across a network of major health systems. The company has increasingly positioned itself as a key infrastructure provider for AI-driven healthcare research and operational intelligence.

As more healthcare AI companies seek access to longitudinal clinical data rather than isolated datasets, partnerships like this may become increasingly important to model development and deployment.

The Future of Behavioral AI in Medicine

Knit Health’s launch highlights a broader evolution in healthcare AI: a transition from systems trained primarily on medical knowledge toward systems trained on institutional behavior.

If successful, this category of behavioral AI could eventually help hospitals standardize high-quality care delivery across large organizations while reducing operational friction that contributes to clinician burnout and delayed treatment.

The approach could also influence how future AI systems are developed in other industries where institutional workflows and human coordination matter as much as formal documentation.

For healthcare specifically, the long-term implications extend beyond automation. Systems capable of learning from millions of real-world patient journeys may eventually help identify operational patterns associated with better outcomes, enabling health systems to continuously refine care delivery based on observed behavior rather than static guidelines alone.