The agentic AI assist Stanford University cancer care staff needed

Every day, thousands of people are diagnosed with cancer around the world. Each case is unique, with hundreds of distinct tumor subtypes that require treatment protocols involving new drugs, clinical trials, and device-based therapies. Leading cancer centers, therefore, rely heavily on multidisciplinary tumor boards, or specialized sessions where radiologists, pathologists, surgeons, oncologists, genetic counselors, and other specialists perform sophisticated analyses of vast amounts of patient data and parameters to develop personalized care plans.  

A recent study by the American Society of Clinical Oncology (ASCO) found physicians spend between 1.5 and 2.5 hours per patient meticulously reviewing images, pathology slides, clinical notes, and genomic data. In this context, agentic AI has extraordinary potential to reduce admin friction and transform how medical services are delivered.

At Microsoft Build 2025 earlier this month, Nigam Shah, CDO for Stanford Health Care, discussed agentic AI’s ability to redefine healthcare, especially in oncology, as physicians get overloaded with the administrative tasks of medicine, he said, which lead to burnout. “Add to this that medical knowledge doubles every 60 or 70 days, so it’s very difficult to keep up with the medical literature,” he added.

Shah also explained that doctors today spend too much time not making medical decisions, something they hope to transform with automation, software development, and agentic AI. “This new era of agents represents an opportunity to begin eliminating some of that undifferentiated work of searching for the right clinical trial, consulting the correct bibliography, and organizing data from different sources and types,” he said. In his opinion, something that makes a difference is the way he and his team have approached software application management within the healthcare ecosystem.

“Most healthcare organizations and companies work with Windows as it’s something they know and trust,” he said. “That’s why we’re committed to building a powerful use case based on this to drive new capabilities through specialized multimodal AI agents.” To do this, they rely on Microsoft and its healthcare agent orchestrator found in the Azure AI Foundry Agent catalog.

Rewriting tomorrow’s healthcare today

The healthcare agent orchestrator includes pre-configured agents, as well as open-source customization options, that enable developers and researchers to coordinate multidisciplinary and multimodal healthcare data workflows, like tumor panels. It also streamlines deployment into healthcare enterprise productivity tools such as Microsoft Teams and Word.

In general, modular reasoners and specialized multimodal AI agents work hand in hand to tackle tasks that would otherwise take hours, effectively complementing clinical specialists with personalized, cutting-edge AI. So by integrating the latest Microsoft capabilities, the healthcare agent orchestrator can handle analysis and reasoning across diverse types of health data, from imaging and pathology to genomic data and electronic health record (EHR) clinical notes. Each agent is equipped with advanced AI models from Azure AI Foundry, which combine general-purpose reasoning capabilities with healthcare modality-specific models to generate actionable insights based on multimodal clinical data.

Stanford Medicine currently serves 4,000 tumor board patients annually, and its physicians already use summaries generated by the base model in meetings using a secure GPT Phi instance in Azure. “The new healthcare agent orchestrator has the ability to streamline this existing workflow by reducing fragmentation and enabling new insights from previously difficult to search data elements, such as clinical trial eligibility criteria, treatment guidelines, and real-world evidence,” Shah said, adding that this is how workloads are shortening, resulting in reduced burnout and overload rates among medical staff caring for cancer patients.  

“Stanford Health Care is excited to continue exploring the potential of the healthcare agent orchestrator to develop the first generative AI agent solution used in a production environment for real-world cancer patient care,” he said.

Accelerating innovations for care teams

As clinical care complexities increase, the healthcare agent orchestrator enables developers to confidently navigate the accelerated era of agentive AI, collaborate with clinicians, and democratize precision medicine tools. The initial framework is designed to support tumor boards, but the ultimate vision is to empower healthcare and life sciences developers in how agentive AI can impact clinicians and patients more broadly, providing real-time support to multidisciplinary care teams across the healthcare ecosystem. 

For Shah, oncology committees are just one example among countless use cases where technology can help. “The most common question doctors ask is what happens to other patients like theirs,” he said. “Now there’s an AI-based agent capable of analyzing a sample, discerning between parameters, and reaching a conclusion about what happened. Not only do we have technology to help us with the heavy lifting, but it’s also capable of providing immense added value on a larger scale.”