In the latest installment of our BrightInsight Digital Health C-Suite Series, we spoke with Emmanuel Frenehard, Chief Digital Officer at Sanofi, about the company’s approach to emerging AI technologies, the concept of “responsible AI,” and the importance of vendors such as BrightInsight in driving the innovation ecosphere. Here’s a recap of the discussion. You can watch the interview here, along with other interviews in the 2024 JPM Digital Health C-suite series here.
The French multinational pharmaceutical and healthcare company, which views AI as a way to accelerate its business on a number of fronts, segments the technology into three categories. The first, "specialized AI," is used by Sanofi researchers to detect anomalies in the manufacturing process. The second, "snackable AI," is an app available to the company’s 10,000 employees that conveys predictions in bite-sized chunks – "think of it as social media and AI."
The third category is generative AI, which the company divides into four areas:
Sanofi has used LLMs in drug discovery for some time, but recent advances have delivered new levels of richness and expanded parameters, enabling more natural exchanges between human and AI. "Drug discovery LLMs, because they contain information that is very unique, and we think very powerful, but also potentially the interface between the human and the technology of gen AI being far more natural, I think far more powerful."
Sanofi’s company-wide Responsible AI initiative "was an absolute necessity."
"We're saying that we will be accountable for it, it has to be fair and ethical, it has to be robust and safe, and last but not least, it has to be eco-friendly, because it can be quite wasteful to let things run in the clouds and out of sight, out of mind."
That means each decision or recommendation driven by AI can be traced and explained. "What we do, even when we partner with companies such as BrightInsight, has to be that we're accountable to the outcome."
That accountability becomes even more important considering the potential for AI "hallucination." A recent study published in the Journal of the American Medical Assn. found that ChatGPT version 4 had a diagnostic error rate of 83 percent in pediatric cases.
"The risk of hallucination is something that we worry about. It's really important that when we engage with patients or we give advice to healthcare professionals, that it is validated and that it is clinically sound." For Sanofi, that means human oversight of AI-generated outputs. "We have to be careful with that, that the human doesn't become too complacent and say, 'Yeah, I'll just validate, because nine times out of 10 the recommendation has been good, today I'm feeling tired, I'll validate and I won't check it.'"
Harnessing AI technologies can shorten the distance between the patient and the treatment. Emmanuel mentioned for example, that for some patients with a rare disease, it can take years, even decades, to get the right diagnosis because they may present with common symptoms, like headaches, not sleeping well, fatigue, etc. "I think artificial intelligence, I think technology, I think the empowerment of all of these capabilities in the hand of a human can accelerate that, can fast track that." Faster, accurate symptom detection and diagnosis can get patients on medications sooner to improve outcomes.
Digital health technologies such as the BrightInsight Platform are driving a major transformation in the way healthcare is delivered. "Democratization of healthcare is happening because of companies like [BrightInsight]." Sanofi views third-party vendors such as BrightInsight "innovation companies," crucial to driving advances in its portfolio. "There's vendors and there's partners, and partners are very different. We think our partners are an extension of us."
"[BrightInsight has] done a lot of things for us that are very innovative. We want to make sure that we jointly look at those problems, because you may have a very different twist on it that we had not thought of, an approach that we had not imagined."