AI In Pharma Moves From Experimentation To Controlled Deployment At Tolmar

Tolmar is applying artificial intelligence (AI) across pharmaceutical development and manufacturing with a focus on productivity, data unification, and risk-based governance, according to Jon Steffey, Senior Director of Enterprise Software and Analytics at the company.

Speaking at PharmaCon 2026, Steffey described AI as a tool to accelerate workflows rather than replace decision-making in regulated environments. “Across almost every industry, AI will be a productivity enhancer,” he said during an Adastra podcast recorded at the event.

He added that certain judgments remain with humans, particularly where patient outcomes are involved, requiring a balance between automation and oversight.

Industry data supports that framing. Pharmaceutical companies are already using AI in small molecule design, simulation, and manufacturing optimization, with deployments reducing development timelines and improving operational efficiency.

At scale, companies such as Pfizer have reported measurable gains in AI-enabled drug development and manufacturing workflows, including reduced discovery timelines and improved throughput.

Data Foundations And Unified Platforms Replace Fragmented Systems

Tolmar’s approach centers on building a unified data platform to address fragmentation across manufacturing, enterprise resource planning (ERP), and laboratory systems.

“When I joined Tolmar, we had good reporting but in silos,” Steffey said. Decision-makers often had to pull data manually from multiple systems to understand manufacturing performance.

The company is consolidating these systems into a single environment using modern data platforms, with an emphasis on ingestion, transformation, and governance rather than isolated use cases.

Steffey said organizations often attempt to deploy AI before addressing these foundational elements. “There is a temptation to chase a single killer use case and skip the foundational work,” he said.

This aligns with broader industry challenges, where fragmented data systems limit AI deployment and decision-making. Healthcare organizations require unified, structured datasets to enable reliable analytics and automation. Efforts to unify data are also underway across the commercial side of pharma. Companies are consolidating customer, clinical, and operational data into single platforms to improve visibility and coordination.

At Tolmar, the unified platform enables an end-to-end view of manufacturing and quality data, which Steffey said reduces uncertainty and improves decision-making.

“Structuring and federating the data… makes decisions more empirically driven,” he said.

Risk-Based Governance Shapes AI Deployment In Regulated Environments

Steffey emphasized that AI adoption in pharma is constrained by regulatory expectations and product risk.

“If it is a cancer drug going into a human body, you cross the t five times. If it is an AI-developed cancer drug, maybe you cross it six or seven times,” he said.

He contrasted this with lower-risk applications such as user interface changes, where development cycles can be shorter and less restrictive.

This risk-based model reflects standard regulatory approaches in pharmaceutical manufacturing, where validation, auditability, and data integrity requirements increase with potential patient impact.

AI is also being applied to digital twins and simulation-driven development, allowing companies to test processes virtually before physical deployment. These systems support predictive quality control and process optimization in regulated environments.

Large-scale investments reflect this shift. Eli Lilly, for example, is building AI-enabled manufacturing infrastructure as part of new production capacity expansion.

Steffey said the primary constraint is not technology but organizational adoption. “Most of the technical challenges… are solved problems. The hard part is how do I bring my users along,” he said.

At Tolmar, the focus remains on aligning AI deployment with business priorities, using data systems to support controlled, measurable improvements in manufacturing and product development.