Cognizant leaders on the tech trends that will define 2025
From industries under pressure to hot growth regions to workforce trends, our experts weigh in.
March 08, 2024
Value-driven, scalable gen AI programs can deliver a lasting competitive advantage for biopharma firms. But only if they move quickly and think differently.
In March 2023, NVIDIA launched BioNeMo™, a cloud service that aimed to accelerate drug research and discovery through large language models and generative AI. At the time, media coverage focused on the platform’s potential—how it might rapidly identify potential drug molecules, predict their interactions in the body, and even simulate some aspects of clinical trials.
A year later, it looks like the hype around BioNeMo may, if anything, have understated its potential. In the words of David M. Reese, CTO of Amgen, an early-access customer of the service, BioNeMo has been a “hinge moment” for his company, which will soon be using the service. It will train state-of-the-art AI models on one of the world’s largest human datasets in a matter of days rather than months. That could mean significant improvements in time to market, particularly impacting drug discovery and phase 1 clinical trials.
Amgen’s proactive embrace of gen AI, in other words, is already bearing fruit, with the prospect of even richer harvests in coming years. For other life sciences companies that want to follow suit, they need to move quickly and think differently. As a starting point, in this post we offer five core truths that companies can embrace to create value-driven, scalable generative AI programs and turn them into a multi-year competitive advantage.
While there is a dizzying number of exciting potential use cases for generative AI in the life sciences arena, it makes sense to focus initial investments on those that either shorten the discovery and development periods or amplify the scalability of treatments post-market.
As every pharmaceutical leader knows, the success of a new drug is a matter of speed and scale. Companies usually have between five and eight years to recoup the discovery and development costs of bringing that new drug to market…and funding the next one. When it comes to gen AI, an adoption strategy that puts speed and scale first is the key to realizing a competitive advantage. On the clinical front, this approach may resemble Amgen’s collaboration with NVIDIA, streamlining discovery and development phases through the application of advanced models that swiftly evaluate molecules, or facilitate in silico clinical trials.
On the commercial side, generative AI is an invaluable tool to help educate healthcare professionals, hospitals, payers, insurance companies and even patients about a treatment’s value, thus helping companies capitalize on the limited in-market growth phase. Efforts here can take the form of microbrand launches, precision targeting of HCPs, optimized contracting and pricing strategies, more personalized customer and patient engagement, and/or better field force effectiveness.
In the end, the main commercial objective should be accelerating access to therapy and care to all eligible patients. That drives better patient outcomes, which in turn drives better business outcomes.
Figure 1
Takeda Pharmaceuticals’ experimental psoriasis drug, which was purchased from U.S. startup Nimbus Therapeutics, is based on a compound that was discovered by an AI-enabled algorithm. The model identified the compound in just 6 months – nearly three times faster than the usual 2-year candidate molecule discovery period.
While gen AI is grabbing headlines in this story, the unsung hero is really the data being used by the model.
As with any system, the old adage of “garbage in, garbage out” applies to gen AI tools. As life sciences companies begin their AI journey, they must take a long, unflinching look their data estate. Do they already have the accurate, complete, timely and relevant data they need to support and scale use cases? Or should they spend time and resources building the solid data foundation so critical to generative AI?
Either way, time is of the essence. Only with robust data “plumbing” in place can companies start reaping competitive advantage—both in the short term, the so-called “experimentation phase,” and the ensuing period of “confident adoption” where that competitive advantage can be widened. We recommend democratizing data access across the life sciences ecosystem through a process that automates data infrastructure and establishes a sturdy governance framework.
If we accept the idea that creating a competitive advantage from gen AI requires companies to act quickly, then it will in most cases rule out the idea of building an LLM from scratch.
The alternative is to use one of many existing public models—but these have their limitations. While an out-of-the-box solution may be the quickest way to jumpstart a gen AI program—especially for commercial applications—that can make it harder to leverage the proprietary data that is key to many advanced use cases.
(Not only that, but there is currently an active debate over whether and to what extent commercial AI models can use copyrighted information in their training process. While the resolution of this legal issue is still pending, companies should view it as a potential downside of so-called “public” LLMs as they’re formulating their gen AI strategies, particularly when relying on public models.)
For many organizations, the optimal solution lies somewhere in the middle: so-called retrieval-augmented generation (RAG) is a way to optimize a commercial LLM with data from a private, authoritative knowledge base. Regardless, if that combination shakes out, the output should have a high degree of transparency regarding the source(s) that were used to generate a given response.
This is where forming partnerships can help an organization start experimenting and exploring right away. In many cases, these partners can help companies fine tune an existing model to meet their specific needs, enabling them to quickly build powerful, customized applications while ensuring the safe and secure use of proprietary data.
It’s fun to speculate about hype cycles, and gen AI may be a boardroom favorite today. But it won’t be long—12-18 months? —before the board will want to see results. It’s important, therefore, for those stewarding the first phase of gen AI adoption to choose use cases that will yield quantifiable results to justify additional investments.
The most logical place for most biopharma companies to start is with commercial use cases that span sales, marketing, customer care and market access. These use cases can offer substantial productivity gains of up to 45%, helping companies build awareness of new drugs without introducing significant regulatory risk.
Figure 2
As discussed earlier, it’s important to organize all use cases within an overarching framework that lets teams consistently demonstrate the value of programs, scale successful initiatives and continuously iterate to drive improvement.
Gen AI is redefining work as we know it. But at least for now, and the foreseeable future, gen AI applications require human oversight—especially in the high-stakes healthcare and pharmaceutical sectors. All eyes are on the steady progress toward artificial general intelligence (AGI)—AI systems that can learn, reason and adapt across domains in a human-like way. But until AGI arrives, if it ever does, gen AI can still not yet be fully trusted to navigate a complex healthcare ecosystem while ensuring safety and delivering an empathetic patient experience. Even with AGI, which might automate narrow diagnostic and quantitative tasks to assist providers, a shift in ultimate accountability for patient outcomes is unlikely. There are also important, complex questions to be answered around ethical boundaries, regulations and controls.
Staying mindful to this issue, the ineradicable human factor of all healthcare, can help companies overcome one of the biggest barriers to effective, at-scale use of gen AI-enabled tools: the willingness of people to use the solution and/or help build the ecosystem needed to maintain the tool.
Another way to help ensure healthy adoption of gen AI tools is to co-create the solution with members of the user group. This ensures that the tool's recommendations are user-friendly and not overly complex, fostering trust as users become acquainted with the technology.
As we find ways to use generative AI to help us work more efficiently and effectively, it has become clear that this technology will transform the biopharma sector. The writing is on the wall and has been for some time. Long before Amgen made its big bet with NVIDIA, there was Novartis CEO Vas Narasimhan’s declaration that the Swiss drugmaker should reimagine itself as a “medicines and data science” company. For its part, Sanofi described itself as “all in” on AI and data science when rolling out its plai app.
These industry leaders are infusing their organizations with new capabilities driven by data and AI to realize tangible competitive advantage right now. Those that lag behind on generative AI risk more than a missed competitive opportunity. They risk watching an entire industry evolve without them.
Want more insights from the author? Watch our recent webinar, A Commercial Accelerator: Generative AI's Most Immediate Impact on BioPharma, hosted by Reuters and Cognizant.
Vyom Bhuta is the Global Head of Commercial Innovation for Life Sciences at Cognizant. He and his team help pharmaceutical, biotechnology and medical device companies commercialize their scientific innovations into products and brands that drive patient and business outcomes.
The Modern Business newsletter delivers monthly insights to help your business adapt, evolve, and respond—as if on intuition