
[vetal1983/Adobe Stock]
Eversana is among those that are taking action. The company recently announced a partnership with AWS to increase adoption of generative AI in life sciences.
Pharma’s generative AI journey: Balancing fear, FOMO and practical experimentation

Scott Snyder
Looking at the landscape, Snyder said that a lot of firms are in the experimentation phase with the technology. Snyder describes the current landscape as a time of experimentation, with companies figuring out their path. “Right now, it’s a lot of experimentation for a couple of companies that have really leaned into certain use cases,” he says. For instance, some companies are using AI for regulatory document processing, while others are exploring automation of adverse event case intake.
Generative AI in particular offers potential to accelerate “content velocity,” Snyder said. Although not error-free, it can both generate and also analyze content at superhuman speeds. Generative AI’s capabilities make it possible to, say, take a piece of educational information on a disease and produce hundreds of versions of it for distinct demographics — supporting a variety of languages and locales. “Maybe that resonates with different age groups or demographics, instead of having to go back to the studio or the design board every time,” Snyder said. “That’s one thing generative AI can do really well — ‘Hey, pick this video, but make it localized for Peru right?'” he said. “And so it might not get you 100% of the way there, but it might allow the designer to say I can finish the last 10%.” Similarly, generative AI tools could cut the review time for medical documents by 50% to 80%, Snyder estimated.
But generative AI offers more than efficiency for niche use case. The technology could fundamentally transform pharma’s operating models, Snyder said. New AI capabilities could mean rethinking everything from patient support centers to sales team structures. Snyder describes workshops that identify high-potential AI use cases for pharma clients and assess how integrating those deployments could transform their operational roadmaps.
Navigating future uncertainties with a flexible approach
To future-proof applications of generative AI, Snyder espouses the idea of flexibility, recognizing the value in a diverse range of models suited for distinct use cases. With the number of generative AI model options quickly evolving, flexibility becomes a crucial consideration. It is important to have a diverse toolkit of models tailored to different uses. By taking a “bring-your-own-model” approach and remaining open to a spectrum of AI tools, Snyder believes companies can equip themselves for an uncertain future. Rather than betting on one horse, he recommends stacking your roster with AI tools that excel at specific functions.
To address the hesitancy of the risk-averse pharma sector, Eversana is first focusing on medical legal review. “We think if we can get that to a level where people trust it, all of a sudden, that takes a major friction point out,” he said. According to Snyder, ethical AI development frameworks could serve as the guiding light. “Every company if they don’t have one should have an ethical AI development framework,” he says, pointing out that these should highlight principles of transparency, reliability, trust, accountability and fairness.
The application of these principles, according to Snyder, leads to a more responsible AI. companies guided by such guidelines will have a measure of responsibility. “You’re going to set up sandboxes,” he said. Generative AI experiments in such environments won’t, say, cause harm to a customer before it is deployed in a production environment.
Uncharted territories: Navigating cost implications on pharma’s generative AI journey
The journey towards AI adoption is not without complexities, Snyder said. One challenge companies face currently is the uncertain cost of AI adoption. There are also options to consider. “Are you just going to pay by the drink? And have employees turn it on? Are you going to build your own models and host them?” he asked.
Snyder provides examples of potential use cases like automating medical regulatory review or accelerating literature searches on behalf of doctors. For each case, Snyder suggests, organizations can estimate time savings and the value of those savings, providing a foundation for understanding the investment required. Snyder underscores the importance of determining, say, “micro-economics of automating medical regulatory review with gen AI.”
Such calculations also allow for comparison of different implementation options. Snyder suggests that building a proprietary model, or using a hybrid public model fine-tuned to the organization’s needs, would come with different costs and risks. Once pharma companies have determined the value of time savings, they can then evaluate the type of investment necessary to achieve that value. The decision could be between building their own models or using a hybrid public model tailored to their needs. Both options, Snyder said, come with “different costs associated with them, and different risks”.
Prioritizing practical applications while staying abreast of advancements
Ultimately, Snyder advises companies to focus the majority of their efforts on immediate, viable applications, while keeping an eye on advances that could potentially disrupt the landscape. “My advice to enterprises like spend about two-thirds of their time on the low-hanging fruit use cases,” moving along with technology vendors who are already well-known in the field, he said. Snyder recommends that the remaining third of a company’s attention should be spent scanning for advancements in generative AI platforms, “because they may emerge for certain kinds of use cases and disrupt the economics or the performance.”
Filed Under: Data science, Industry 4.0, machine learning and AI