Pharma SOS Learnings: The Future of AI in Pharma

Earlier this month, I had the opportunity to participate in a panel discussion on artificial intelligence and machine learning at the Pharma SOS 2024 conference. The conversation resulted in valuable guidance for life sciences companies on where AI can make the most impact. Here are my takeaways:

  1. AI will be highly valuable for decision-making. Predictive AI reduces the cost of prediction. Generative AI reduces the cost of creation. By handling the tasks (and associated costs) of prediction and creation, AI tools empower users to make decisions and take creative actions in their workflows.
  2. Companies must implement a fit-for-purpose approach. A user’s workflow encompasses a wide range of decisions and creative actions. However, there are one or two core decisions and creative actions that drive the rest of the workflow. Therefore, companies must pursue “fit-for-purpose” solutions rather than “fit-for-everything” solutions. With a “fit-for-purpose” solution, AI addresses specific challenges (such as improving the response rate to digital marketing materials among target hematologists), rather than high-level goals like growing market share.
  3. Combine structured and unstructured data for impact. In the short run, I believe companies will get the most value from generative AI by effectively marrying unstructured proprietary data with structured data. This will provide richer insights for better strategy, customization, training, and more.
  4. Prioritize domain-specific data. Smaller LLM models trained on domain-specific data deliver better performance in terms of latency, accuracy, and cost compared to larger foundational models. This also creates a flywheel effect, which helps to further develop a competitive advantage.
  5. Domain knowledge + AI will provide a competitive edge. The vast majority of really good AI model engineers and data scientists focus on building models that deliver at scale and scope. This is an advantage, but the life sciences industry needs to develop deep empathy for customer pain points, where modelers and domain experts will be required. This is a tech + domain marriage.

Thank you to my co-panelists for a lively and interesting discussion: Shan Yeh from Jazz Pharmaceuticals, Brian Garino from Novartis, Murali Pinnaka from Arcutis Biotherapeutics, Igor Rudychev, president of the Pharmaceutical Management Science Association, and Prakash Karaka from Chryselys.

If you’re exploring AI opportunities, wondering where AI can make the biggest impact in your company, or struggling with the foundational data and technology infrastructure to leverage AI tools, let’s talk!