Skip to main content Skip to footer
Case study

The challenge

A national healthcare agency (HA) had identified the potential for generative AI to provide medical advice to patients with non-emergency symptoms. It envisioned a human-like avatar that would use large language models (LLMs) to triage symptoms and recommend a course of action. 

This ‘virtual clinician’ would be capable of listening to patients describe their symptoms and asking the same kinds of clarification questions that a human doctor would ask. Using knowledge absorbed from reams of peer-reviewed medical literature, it would arrive at a diagnosis and advise the patient on what to do next.

There was only one problem: the HA in question didn’t have the expertise internally to build a sophisticated generative AI clinician. It turned to Cognizant’s global AI and Analytics team for help.

Our approach

Our initial discussions with the HA clarified that the vision was to use an animated AI avatar to replicate the clinical pathways that real-world doctors take to drive diagnosis and suggest next steps. To orchestrate that vision, we designed a generative AI solution that works at four interconnected layers: intent, information, cognition and presentation.

Working closely with clinical specialists at the HA to define real-world triaging questions, including counterfactual questions to minimize misdiagnoses, we built a working virtual AI clinician in just three weeks. The project phases over that period incorporated ideation and scoping, building and layering the LLMs, refining the models based on feedback from clinicians, and beta testing with internal stakeholders and the launch to external testers.

A window into the future of telemedicine

The virtual AI clinician has now been proven as a single, scalable solution with built-in clinical governance, capable of diagnosing over 900 common medical concerns across an entire population. Designed, built, tested and launched in just a few weeks, the solution provides an efficient, accurate and cost-effective service for first-line medical care. 

The solution offers a glimpse of the future of telemedicine and digital therapeutics. It can help resolve multiple operational challenges facing modern healthcare systems—for example by alleviating pressure on healthcare contact centers, removing some of the diagnostic burden from busy clinicians and enabling patients to get rapid, accurate and science-based advice for a wide range of symptoms. 


individual conditions capable of being triaged 


patient conversations handled during testing phases 


accuracy of AI-powered diagnoses