AI Talk: AI in the pharmaceutical industry – is your data a strategic asset or a hidden risk?  

Views on news 

Combined pressures make it essential for life sciences companies to find new ways to boost growth, improve productivity, and increase operational agility. In pharma, 75 to 85 percent of workflows contain tasks that could be enhanced or automated by agents, potentially freeing up 25 to 40 percent of an organization’s capacity. In medtech, the figure is 70 to 80 percent.

 

These capacities are at the task level, so they may be fractions of employees’ time. McKinsey’s survey found that the opportunities for agentic AI in the pharmaceutical and medtech sectors are similar in most domains, except in R&D, where their approaches differ significantly. 

 

To see tangible growth from AI deployments, businesses must put more effort into managing change, upskilling and reskilling the workforce, redefining business processes into tasks that can integrate with agents, as well as data security and keeping a human in the loop. The survey also found that through a 7-9% time investment in partnering with AI, a firm can get a 40-50% improvement in productivity. The key to the right approach is treating AI as a colleague and training it to understand workflows alongside with the realisation that – despite the hype – it’s an evolution.  

 

How agentic AI is reshaping pharma 

There has been no technology yet that has been adopted by commercial users as quickly as gen AI but the technology’s long term potential in business is still underestimated. AI should increasingly reach beyond IT departments to show the value of AI solutions across the whole enterprise. MCP servers acting as intermediaries between AI models and internal systems like databases, source code managers, SaaS apps and the like are secure interface layers connecting models to enterprise resources.

 

It’s the MCP protocol’s role too to limit the scope of data the AI model looks into. AI in R&D can be leveraged for identifying patient populations, simulating trial outcomes and for partner/vendor CRO section. For AI deployments to enhance executive level decision making it’s still early days. There are, however, more opportunities in clinical trial design, site and country selection and patient population design.  

 

Success will depend on the company’s data assets and how they get organised. Deployment should start with identifying individual workflows that can be carried out by agents and then bringing all the different agents in line through orchestration. Taking trials as an example, there can be an agent for recruiting patients, a data monitoring, a regulatory compliance agent and ones tasked with document processing and monitoring sensor and device data.  

 

Cegeka’s major agentic AI project, for example, has been around large scale pharmaceutical documentation, using GraphRAG to provide context for the LLM by connecting proprietary systems with external sources. Clinical protocol development is another area, where AI agents can identify similar trials and exclusion criteria and standard of care to those prevailing in a pharmaceutical and leverage those to generate documents. AI can also pull vendor data and compare prices and the expertise of the clinical staff, as well as identify inconsistencies. Replacing reactive models with proactive planning and forecasting and streamlining document retrieval with AI agents can save a pharmaceutical millions of dollars.  

 

The panel’s advice 

• Start your journey slow with some low-complexity agents and start experimenting from there.  
• Pick a project that’s important for you and start a pilot.  
• Implementing AI agents is not just about translating human jobs into automated workflows but also getting jobs done that humans couldn’t do because they were either overwhelming or weren’t commercially viable.  
• The preliminary steps to AI deployment are creating an AI governance structure, finding dedicated people for oversight, addressing the issue of human responsibility and putting monitoring in place for these factors. 
• AN FDA guidance document on the use of AI in clinical trials was published in January 2025.