AI Is Changing How We Design Medical Devices

I want to be precise about what I mean when I say AI is changing medical device design, because the claim gets made loosely. This is not about AI writing design briefs or generating concept sketches. It is about AI changing the actual evidence base that design decisions are made from.

That distinction matters. Design decisions in medtech are high stakes. A poor ergonomic decision on an insulin pen can lead to dosing errors. A confusing interface on a defibrillator can cost seconds that matter. AI is beginning to change how quickly and accurately we can identify those failure points before they reach a patient.

Where AI Is Actually Making a Difference

The most useful application I have seen is in user research synthesis. A comprehensive usability study for a medical device generates hundreds of hours of observation data. Identifying patterns manually is slow and prone to analyst bias. AI tools can now process that data faster and surface risk clusters that a human reviewer might miss on a first pass.

The second major area is predictive failure modeling. Traditional DFMEA work is table-driven and largely based on expert intuition. AI-assisted FMEA tools are beginning to cross-reference field complaint databases, adverse event reports, and simulation data to generate more complete risk profiles earlier in development.

What We Are Doing at Dip Studio

When GE HealthCare brought us in on their ecosystem work, one of the things we focused on was how design decisions propagate through a connected device system. A change to an interface element in one device creates downstream effects on how clinicians interpret outputs from other devices in the same workflow. AI-assisted simulation is making it possible to model those second-order effects before anything is built.

We are also using AI tools in image analysis during formative usability testing. Instead of manually reviewing hours of eye-tracking footage, we can identify attention failures much faster. That compresses the iteration cycle in ways that have real commercial value. Faster iteration means faster regulatory submission timelines.

What AI Cannot Do

The design decisions that require judgment about human dignity, about patient experience, about what a device communicates to someone who is frightened and in pain, those still require human designers who understand those experiences. AI can surface patterns. It cannot tell you what a chronic disease patient feels when they look at a device they will wear for the rest of their life.

The firms that will use AI well are the ones that treat it as an input to human judgment, not a replacement for it. The firms that will struggle are the ones that assume AI output equals design quality.

The Regulatory Timeline Is Changing

The FDA is actively updating its framework for AI-enabled devices. The 2024 action plan for AI and machine learning in medical devices set a precedent that manufacturers need to understand now. The devices being designed today will be submitted into a regulatory environment that will be materially different from the one that existed three years ago.

Studios and manufacturers that are building AI fluency now, in their design processes, in their documentation, in their testing methodologies, will be better positioned to navigate that environment. This is not a distant technology concern. It is a business readiness question for right now.