BLOG
Generative AI (GenAI) has rapidly gained traction across industries—but what’s the return so far? Is it truly making a measurable difference for businesses, and how does that translate into financial value? A recent McKinsey report highlights a paradox: while around eight out of ten companies say they’re using GenAI, just as many report no meaningful impact on their bottom line—a phenomenon the firm dubs the “GenAI paradox.”
The life sciences sector shows even higher adoption rates than most industries, yet it too is still waiting on significant returns. In a McKinsey survey of 100 pharmaceutical and medtech executives from earlier this year, only 5% said GenAI has become a competitive differentiator delivering consistent, significant financial value. Still, optimism remains high: over two-thirds of respondents said their organization plans to increase GenAI investments.
As a way to break out of the GenAI paradox, McKinsey suggests a shift to AI agents. These agents have the potential to swing GenAI into a more proactive, virtual collaborator motivated to accomplish goals. This shift would ultimately improve not only efficiency and agility but open up avenues to new revenue opportunities.
Agentic AI combines the raw power of large language models (LLMs) with the ability to plan, orchestrate, and complete tasks by interacting with other tools, data sources, and services. Rather than a single intelligent assistant, AI agents work as “a team” of specialized experts, collaborating behind the scenes to complete an end goal.
Looking for an introductory session on agentic AI for the life sciences? Learn more about the technology, its capabilities, and its requirements in this ONTOFORCE webinar recording. We share insight into how the technology works (with practical demonstrations), and advice on how to prepare your organization’s technology ecosystem. Watch now >>>
The true power of agentic AI rests in its ability to change up workflows and established processes to potentially make them less costly or risky. For life sciences organizations, that might look like streamlining regulatory submissions by automatically gathering, interpreting, and formatting data across clinical, safety, and quality systems. Or it could mean accelerating drug discovery by having agents autonomously generate hypotheses, search relevant databases, and design preliminary experiments. In clinical trials, agents might coordinate recruitment logistics, monitor site compliance, and synthesize trial data in real time to inform adaptive study designs. These aren’t just efficiency gains, they represent a shift toward more agile, insight-driven operations that scale with complexity.
.png?width=700&height=393&name=Watch%20ONTOFORCES%20agentic%20AI%20webinar%20(1).png)
Gain insight into how agentic AI works (with practical demonstrations) and advice on how to prepare your organization’s technology ecosystem. Watch now >>>
Adopting agentic AI for life sciences companies requires both tackling technical and human challenges. From a technical perspective, an organization’s entire AI approach might need rethinking: moving away from scattered initiatives and pilots into a deliberate strategy that enables AI to power processes and eventually new business capabilities.
From a human perspective, working successfully with AI agents requires trust. It seems that the life sciences industry has already partially tackled this hurdle. A 2024 Cape Gemini report details that the industry leads in agent adoption and that many industry execs (63% of those interviewed) would trust AI agents to analyze and synthesize data. Further, half said they would trust an AI agent to send a professional email on their behalf.
There are also broader concerns that may slow adoption, particularly around data privacy and explainability. Life sciences companies handle sensitive patient and clinical data, making regulatory compliance and security non-negotiable. Further, giving AI agents autonomy in decision-making raises flags about transparency and accountability, especially when those decisions can potentially influence trial outcomes, treatment paths, or regulatory submissions.
One way to address these challenges is by integrating knowledge graphs. These structured, interpretable data frameworks help agents reason over validated, traceable sources, improving both transparency and compliance. By grounding agent decisions in trusted, auditable knowledge, organizations can reduce the “black box” problem and build trust both internally and externally.
Agentic AI is moving beyond early experimentation as more organizations explore its potential and start to see real results. Researchers, along with a growing number of companies are showing that AI agents can do more than complete isolated tasks. Agents can support broader workflows across both research and business functions.
Let’s explore some emerging use cases that highlight how agentic AI is already being used to streamline multifaceted processes in life sciences.
Agentic AI powering personalized medicine processes
A group at Cornell University recently shared their outcomes with an agentic AI project: TxAgent, an AI agent leveraging multi-step reasoning and real-time knowledge retrieval across a toolbox of over 200 tools to analyze drug interactions, spot potential risks, and create tailored treatment plans. It looks at how drugs work on a molecular and clinical level, checks for conflicts based on a patient’s health conditions and medications, and fine-tunes recommendations through reasoning.
TxAgent chooses the right tools for each task based on objectives, runs structured queries, and pulls data from trusted sources. With 92.1% accuracy, it delivers safe, reliable recommendations that follow clinical guidelines and help avoid harmful drug interactions.
Nanobody design – agentic AI to power therapeutic antibody development
In one study, researchers implemented a virtual lab to address an open-ended research problem that required reasoning across biology and computer science. The virtual lab consisted of an LLM principal investigator that guided a team of LLM agents, such as a chemist and a computer scientist. A human researcher was also available to provide feedback. Through this virtual lab, the agents worked together to design nanobody binders to variants of SARS (a viral respiratory disease).
The study reports that the virtual lab created a nanobody design pipeline that utilized advanced tools to ultimately design 92 new nanobodies. Validation testing revealed that many of these nanobodies had promising binding profiles across SARS variants. Notably, two of them bind even better to newer variants, while still working well on the original virus.
Predicting illness spikes to improve awareness – agentic AI powers marketing processes
Bayer built an agent that uses predictive AI and advanced analytics to predict flu trends, improving how products could be marketed to alleviate symptoms in time. Bayer’s consumer health marketing team utilized Google trends data, open-source weather data, and real-time temperature and public flu report data to build a predictive model that could show where and when searches would grow and fall. From there, automated tasks were initiated to optimize Bayer’s Google campaigns, such as adding new, relevant keywords and adapting ad copy. These automations allowed the team to proactively adapt their strategy so they could reach people with the right products to alleviate their flu symptoms.
This agent has helped Bayer’s marketing team in Australia plan and activate more effective campaigns. On top of this, it also has the capacity to scale across departments to open up new capabilities for different functions. For example, Bayer’s product supply team has explored how the agent could help with distribution models during peak times.
Agentic AI is reshaping how life sciences companies explore, decide, and act. From personalized treatment planning to drug discovery and dynamic marketing, early use cases show how AI agents can drive both scientific and operational success. But realizing this potential requires more than experimentation. As McKinsey concludes in their report, agentic AI is not just a stepping stone, it’s the foundation of “the next-generation operating model.” Organizations needs to consider how they can rewire and redefine to fully capture the technology’s potential.
