AI Interface Design

The Future of AI Interfaces in Healthcare

Katina Sostmann and Sergiy Sumnikov explore what comes after the healthcare chatbot: when to use voice, visual, and tactile interfaces; how to keep clinicians and patients in control; and how privacy and trust shape useful AI experiences.

Watch on YouTube Get the transcript

Health innovation designers

Choose the right interaction channel for a patient's or clinician's immediate need.

Clinical teams

Keep practical judgment and human contact at the center of AI-assisted care.

AI product leaders

Balance new interface ideas with privacy, transparency, and real-world use.

Meet the expert

Katina Sostmann

Katina Sostmann is an Associate Partner Health Platform and Design Principal at IBM. Based in Berlin, she has worked in product and service design for more than 25 years, with a focus on digital healthcare. Her experience as a nurse during her studies informs her approach to designing around patients and care teams.

Hosted by Sergiy Sumnikov
General Manager, Health Innovation at Halo Lab

Start with people's daily context rather than an impressive interface. A person managing a chronic condition may need a timely warning, not another dashboard or reminder. Show only the information needed for the decision at hand.

A chatbot is only one form of AI interaction. Voice can support conversation, but a fast visual cue may work better in an intensive care unit, while touch or physical controls may suit other moments. Combine channels only when each one adds value.

Decide where the clinician or patient makes the final call and where AI should step aside. Pair measurements with human observation, and build small opportunities to use judgment so reliance on machines does not erode practical skills.

AI can support a shift from passive receipt of care to active participation. Help people recognize what is happening, understand their options, and build confidence in reading their own health rather than making them dependent on a device.

Treat sensitive health data as a design constraint. Use appropriate environments for different kinds of data, involve patients and other stakeholders in testing, and show users the sources and reasoning behind AI-supported outputs.