Tom Williams | Senior Manager, Product Innovation
Elsevier
Dr. Tom Williams, MBA, is Senior Manager, Product Innovation at Elsevier, where he works on the development and commercialization of scientific connectors and task-level agents for drug R&D. He works with pharmaceutical and biotechnology organizations to translate scientific workflows, evidence requirements, and enterprise integration needs into reusable, machine-consumable capabilities. Tom has a scientific background in biophysical chemistry and neurodegeneration and works at the intersection of scientific information and agentic AI.
Appearances:
BioTechX USA Day 1 @ 12:05
Beyond the Chatbot: Building Trusted Agentic AI with Scientific Services and Connectors for Drug R&D
Pharmaceutical R&D organizations are investing heavily in agentic AI while also developing their own research platforms, copilots, and orchestration environments. However, many implementations still depend on fragmented data sources, inconsistent retrieval methods, and outputs that scientists must manually reconcile before they can support research decisions. This limits reproducibility, makes provenance difficult to preserve, and creates uncertainty about what evidence an AI system has used.
We describe Elsevier’s Life Sciences evolution from a cross-portfolio, multi-agent alpha to a modular architecture based on reusable scientific connectors and task-level agents. Connectors provide controlled access to trusted scientific data and capabilities including retrieve, resolve, normalize structured and unstructured evidence. Agents combine these components to address defined research tasks within broader workflows.
Examples from current work in chemistry, pharmacokinetics, and drug safety include compound identity resolution, chemical structure search, cross-source evidence retrieval, data harmonization, and structured evidence packaging. Separating these scientific capabilities from higher-level agent behavior makes them reusable across different products, models, and customer environments, while allowing provenance, completeness, and individual components to be evaluated explicitly.
Taken together, these examples provide a practical framework for deciding what should be implemented as a connector and task-level agent, or user-facing workflow, and for designing scientific AI that complements rather than replaces the enterprise research platforms pharmaceutical organizations are already building.
Get involved at BioTechX USA
Created by
Terrapinn is proud to be a member of isla.
Working together to build sustainable events