Artificial intelligence (AI) advancements in the medtech industry have the potential to dramatically improve healthcare and patient outcomes. However, the future of AI in medtech remains uncertain. With new regulations like the AI Act set to take effect in 2027 for high-risk devices, many medtech companies are unsure how to use AI to improve their processes while remaining compliant. “Certain things can be sped up enormously by applying AI to them, but it’s not an end-to-end solution,” says Dr. Jochen Tham, head of digital customer experience at Zeiss. “The way AI will do things is different from the way you have learned in the past.”
The AI Act contributes to what Erik Vollebregt, a partner at medtech-focused law firm Axon, calls ‘regulatory lasagna.’ “In medtech, we are used to working with a degree of regulatory lasagna when regulations like MDR and the IVDR converge with other horizontal legislation like the Machinery Regulation or EMC Directive. But, the AI Act adds another layer of complexity,” explains Vollebregt.
For example, many medtech companies will likely have to undergo multiple conformity assessments and obtain parallel CE markings, slowing down an already lengthy and costly process. Plus, the phased adoption of the AI Act will create additional challenges. “Right now, we’re still dealing with the MDR and the IVDR being phased in. When you add in the AI Act, it becomes extremely challenging to develop and launch a product that will comply with the regulations as they continue to evolve,” he says.
Despite regulatory complexity, MDR and AI Act requirements converge in some areas. James Dewar is the Co-founder and CEO of Scarlet, the only EU notified body that specializes in software and AI. Dewar sees real-world evidence (RWE) as a way for medtechs to satisfy post-market surveillance requirements in both regulations while improving product safety and efficacy.
“RWE is derived from real-world data, which monitors how your device is performing in the real world,” says Dewar. “Once you analyze and annotate the real-world data, it turns into RWE.” RWE can satisfy multiple AI Act requirements. For example, companies can plan a RWE study as part of their conformity assessment to show that their device remains safe and effective long-term.
The AI Act also includes an assessment of fundamental rights, where companies must show that they have taken measures to mitigate risks that an AI system may pose on those who use it. “By monitoring your device in a real-world setting, you’ll be able to get data on the specific populations and subgroups that are actually using your device. This will help you understand risks like bias in a much more granular way,” Dewar explains. RWE can also help speed up certification lead times and AI feedback loops when retraining models based on new data or issue detection. “Being able to rapidly detect, collect, and analyze data from the real world is the key to building great AI products,” says Dewar.
To learn more about how Axon and Scarlet are preparing for the AI Act, watch their full Summit session on Veeva Connect.Although the future of AI in medtech remains uncertain, companies are already leveraging AI applications successfully. The key is to invest in a digital platform that lays a strong data foundation for AI tools. For example, Zeiss leveraged its data in Veeva PromoMats to build a proof-of-concept generative AI-based platform to simplify content creation and personalization.
“At the outset, we thought AI would fix our content generation. While it will fix certain steps, like improving prompting capabilities or generating versions, it will not fix your end-to-end content journey,” says Tham. “You still have to make sure that communications are calibrated across countries and disease areas. No one wants to open an email that is addressed to all doctors in the world.”
The overall aim of Zeiss’ AI tool is to help regional marketers quickly create localized marketing assets. “Now, they don’t have to go into a million tools. They have one platform where they can create tactical campaigns to sell to their audience and address local needs while maintaining proper branding,” Tham explains.
Ultimately, Tham recommends that medtech companies think more holistically when considering AI initiatives. This means taking a process engineering view that articulates and solves for business benefits within discrete steps, and then understanding how optimizing those steps will improve the end-to-end process. “There’s this perception that AI is magical and you can sprinkle it on any complicated business process and fix everything,” he says. “Instead, start with a business problem or objective to optimize for. Don’t just say ‘we need some AI.”
To learn more about how ZEISS is approaching AI initiatives, watch their full Summit session on Veeva Connect.