Usability of AI-Enabled Medical Devices
Beyond a Good User Interface
Jouni Karkinen
11/17/20263 min read


Usability of AI-Enabled Medical Devices: Beyond a Good User Interface
Usability engineering is not just about making a medical device easy to use. Its primary purpose is to reduce use-related risks and support safe and effective use.
This is already familiar to medical device manufacturers under the MDR, IVDR and IEC 62366-1. However, AI-enabled medical devices bring additional challenges. Even when an AI system performs well technically, its output may be misunderstood or used incorrectly.
A new IEC project, IEC TS 62366-3, is under development to provide guidance on applying usability engineering to AI-enabled medical devices. It is currently at Committee Draft (CD) stage and is not yet a published standard. TS means Technical Specification. The project is intended to provide guidance rather than introduce additional requirements.
This article focuses on AI-related usability issues that affect interaction between the user and the medical device. AI functions operating entirely in the background without affecting that interaction are outside this discussion.
Three questions worth asking
1. Does the user understand the AI output?
AI-generated outputs may be misunderstood if users do not understand their meaning, limitations or uncertainty. Usability engineering should address how the interface supports correct interpretation and appropriate user actions.
Transparency and explainability are particularly important when the user needs to understand the basis and limitations of AI-generated information.
2. Does the user trust the AI appropriately?
Excessive trust may lead users to accept incorrect recommendations without sufficient consideration. Too little trust may lead them to ignore useful information.
The challenge is to design an interface that supports appropriate reliance on AI, including recognising when a result is unreliable or requires further attention.
3. How does AI change the user's role in the clinical workflow?
AI can change how tasks are shared between the medical device and its user. The Degree of Autonomy (DOA) describes how independently a system performs its functions.
As autonomy increases, the user's role may shift from directly operating the device towards monitoring, supervising and intervening when necessary. Usability engineering must consider whether users understand what the system does independently and when intervention is needed.
Performance validation is not usability validation
AI performance validation and usability validation serve different purposes.
A technically well-performing AI system does not by itself demonstrate that users can interact with the medical device safely. Usability evaluation remains part of the established IEC 62366-1 process, while AI performance validation addresses a different set of questions.
Usability validation testing should involve representative users performing relevant tasks under sufficiently realistic conditions. Particular attention should be paid to situations where misunderstanding an AI output or failing to act appropriately could lead to harm.
The FDA perspective
The FDA's updated August 2026 guidance, Applying Human Factors and Usability Engineering to Medical Devices, also takes a risk-based approach.
It emphasises identifying critical user tasks, reducing use-related risks and conducting human factors validation testing with representative users under sufficiently realistic conditions. These principles apply to medical devices generally and are equally relevant when AI changes how users interact with a device.
What should manufacturers do?
For manufacturers and their development partners, usability engineering should start early, when user tasks, workflows and interfaces can still be influenced by design decisions.
The essential question is not only whether AI produces the correct output, but also whether the intended user can understand and act on that output safely.
Follow the development of IEC TS 62366-3.
A well-performing AI system is not automatically a safe-to-use medical device.
Contact me for any questions.
The views and interpretations expressed in this article are my own and do not represent the official position of any company or organisation.
Sources
Regulatory & Quality Insights
Articles and professional perspectives by Jouni Karkinen.
QA/RA consulting engagements are arranged through Innokas Medical. Scilife eQMS and other enquiries are handled by Karkinen Consulting Oy.
contact information
Karkinen Consulting Oy, Helsinki, Finland
Business ID: 3103786-9
info@karkinen.com
NATO NCAGE code A10CG
© 2026. All rights reserved.


