Clarkston’s Anna Ivashko recently attended the 2026 ISPE AI in Life Sciences Summit in Boston, where industry leaders examined how artificial intelligence is driving change across pharmaceutical and biotech organizations. Conversations centered on practical and scalable applications of AI throughout research, development, manufacturing, and quality, as well as the evolving governance and operational considerations shaping successful adoption.  

The event also provided valuable insight into the strategic priorities and emerging expectations from regulatory bodies influencing the future of AI in the life sciences sector. Across sessions, speakers explored how organizations can move AI from isolated pilots into sustainable, compliant ways of working. Below, Anna shares her key takeaways from the Summit and dives further into what these themes mean for life sciences organizations. 

Key Takeaways: AI in Life Sciences 

1. From Data Science to Human Science 

AbbVie’s Michael Grischeau said it best: “We’re moving from data science to human science.” While the Summit introduced various use cases for AI application — from risk-based audit prioritization to regulatory change summaries and submission support — the common thread from session to session was the humans involved. 

Panelists described key processes to upskill business users on technology and drive adoption through both top-down leadership and bottom-up crowdsourcing. Change management came up again and again, from leveraging change champion networks to considering who to bring into the fold, whether early adopters or loud detractors. Speakers also discussed creative ways to encourage first-time use, along with communities of practice, to facilitate knowledge sharing and bottom-up idea crowdsourcing workshops. 

At the heart of it all is building end-user trust and transparency into these new tools and aligning on the role of the tools to augment, not replace, processes. For life sciences organizations, that trust is especially important as AI becomes more embedded in regulated ways of working. 

As the conversation turned executional, an undercurrent of user adoption remained. Industry leaders recommended avoiding spending all the time on pipeline development before getting to data uses and quick wins. Part of the reason is that those data uses help facilitate adoption and drive momentum behind AI initiatives, while also allowing organizations to differentiate between true value and hype. 

2. Human in the Loop 

Human in the loop was a key topic of discussion throughout the Summit. From using human labeling to train models, such as AI vision capabilities used to support visual inspections, to ensuring AI is complementing, not replacing, conventional processes, it’s clear from both the industry and regulatory body perspective that human judgment will remain engaged.  

Eli Lilly’s Lori Otto addressed this well in her talk by discussing how the organization ensures humans in the loop don’t have automation bias or overreliance on technology for GxP decision-making. Training on risks and limitations of the tools is paramount, along with retraining as the technology evolves quickly. A clear ability for humans to stop or override a recommendation is critical to maintaining control. Evidence in the systems that humans verified information is also key to ensuring compliance.  

However, technology players warned the industry not to allow human in the loop to become an excuse for accepting mediocre tools with poor performance accuracy. Despite this safeguard, they pushed peers to adopt approaches for improving tool accuracy, including targeted models built to support validation and considerations around real vs. synthetic defect usage for training.  

3. Explainable AI 

Sessions across the Summit discussed that AI trustworthiness narratives are necessary for auditors, governance boards, and other stakeholders. At the heart of this is making AI explainable within and outside the organization. 

When it comes to common applications in the manufacturing space, like AI vision, meeting this requirement is less challenging because pictures or videos exist that can be pulled and reviewed after the fact. Even so, most organizations must make a decision around what type of data to store, such as isolating anomalies, and what time horizon of data storage is appropriate. This helps avoid storing everything while controlling data volumes and costs. 

The question all organizations must answer is not just, “Does the model work?” Organizations also need to be able to defend how it was trained, how it’s monitored, and how it’s changing. 

LT Seneca D. Toms (FDA Office of Inspections and Investigations) shared that “Confidence in a system is not established once and assumed forever. It must be evaluated throughout the system lifecycle.”  

4. Data and Process Readiness 

Many panels discussed that laying AI solutions over non-standard or broken processes, or islands of siloed, inconsistent data, isn’t a scalable path forward. Moving down the path of scalability for existing AI pilots requires a holistic strategy to tackle process and data readiness.  

Vivian Huynh of Rockwell captured the challenge clearly: “When we don’t have a digital strategy for automation, we’ll end up with islands of automation. If we don’t have a harmonized data strategy, we’ll end up with islands of AI.” 

Organizations find the largest benefit when they apply AI not to futuristic, flashy use cases, but to very practical, discrete opportunities to remove bottlenecks in quality and manufacturing operations where data and process are both ready. These opportunities may be less eye-catching than broader AI concepts, but they are often where organizations can create meaningful value and build momentum. 

Not every use case requires the same level of control, and the level of control should match the intended use. This can also lift some of the barriers of entry to create immediate quick wins. Once those are underway, knowledge sharing allows for repurposing across larger organizations. Paula Gamboa of Abbvie shared, “Understanding the patterns in the use cases within your organizations is what will allow you to stop reinventing the wheel and start reusing.”  

Looking Ahead: Data Science to Human Science 

The 2026 ISPE AI in Life Sciences Summit reinforced that AI is already being applied across practical areas of the life sciences sector – but the path to scalable adoption is not only technical. 

Across sessions, the common thread was the humans involved. Organizations need to build end-user trust, maintain human judgment, make AI explainable, and address the process and data readiness required to scale. As AI continues to evolve, life sciences organizations that focus on practical use cases, transparent governance, and adoption from the business will be better positioned to separate true value from hype. To continue the conversation, reach out to our team today or visit our AI consulting page.