Artificial intelligence (AI) may dominate discussions across scientific R&D, but successful AI initiatives depend on solving much older foundational problems first: accessible data, connected workflows, clear governance, and organisational alignment.
That was the central message from a recent panel discussion featuring digital transformation leaders from Pfizer, AbbVie, Bristol Myers Squibb, and Revvity Signals. The panel explored the practical realities of digitalisation, data readiness, and AI adoption, while audience polling showed that organisations are moving at very different speeds.
Digitalisation remains a work in progress
The first audience poll asked a simple question: Where are you in your digitalisation journey?
Responses revealed uneven progress. Nearly 30% of respondents reported having automated workflows in place, while an equal share said they had not yet started. Others were evaluating options, running pilots, or working from strategies still awaiting implementation.
Jordan Stobaugh, Director of Scientific Architecture, CMC at AbbVie, described a foundation-first approach. His organisation began by improving cross-functional data sharing, then expanded into broader initiatives supporting collaboration, visualization, modelling, and AI-enablement.
His advice was to start small and build: define the minimum viable data and workflows needed to deliver value, then expand iteratively.
This theme recurred throughout the discussion. Digitalisation is not a single transformation event. It is an ongoing process that must adapt continuously to evolving scientific and business demands
The biggest barriers are not just technical
When attendees were asked about the biggest obstacle to digitalisation within their organisations, the responses were surprisingly balanced.
Lack of prioritisation or strategy, data silos and integration challenges received the most responses, closely followed by legacy systems and scientists’ adoption or change management.
The results showed that digitalisation is not only a technology challenge. Organisational alignment, governance, and change management are just as critical.
Panelists emphasised that leadership support and alignment between scientific and IT objectives are often as important as technology.
David Gosalvez, Chief Strategy Officer at Revvity Signals, warned against overengineering solutions before value is proven. "People want perfection upfront," he said. "The reality is you can add a lot of value with small incremental steps."
Bo Du, involved in Pfizer's Scientific Data Cloud initiative, pointed to data silos and legacy systems as persistent barriers. Decades of scientific data often sit across multiple systems, formats, and repositories, making integration difficult.
The consensus: resist the temptation to "boil the ocean." Focus on high-value use cases, deliver early wins, and build momentum from proven successes.
Metadata and context are the new competitive advantage
As organisations prepare for AI, the conversation around data is shifting from digitisation to context.
Brian Claus, Scientific Senior Director of Cheminformatics at Bristol Myers Squibb, argued that metadata has become as important as the data itself. Information must be understandable to scientists and AI systems. "Our biggest challenge is how do we format our data in some way that the machines can also take as much advantage of it as the people can," he explained.
Du highlighted FAIR principles and the role of metadata in preserving provenance, lineage, and scientific context.
Gosalvez added that metadata captured at the time of experimentation is often impossible to recreate later. The earlier organisations capture context, the more valuable their data becomes for future analysis and AI.
Data readiness is therefore less about where information is stored and more about whether its scientific meaning has been preserved.
AI adoption is accelerating, but foundations come first
The final audience poll focused on AI strategy.
Only 12% of respondents reported actively deploying AI in scientific workflows. The largest group, about 30%, said they were focused on building foundational data infrastructure first. Others were evaluating use cases, running proofs of concept, or defining their broader AI approach.
Claus described efforts to create an environment where scientists can work alongside AI-powered tools, supported by upskilling and rapid experimentation.
Pfizer has expanded AI initiatives across document generation, predictive modelling, drug discovery, and laboratory automation. Still, Du emphasised that implementation remains constrained by governance, regulation, talent availability, and data accessibility.
Stobaugh highlighted report generation and documentation as near-term opportunities that add real value by freeing up scientists' time while organisations continue strengthening their data foundations.
The strongest message was that AI should not be treated separately from digitalisation. Organisations making progress use AI as an extension of broader transformation efforts, not as a standalone technology deployment.
Building the foundation for the future
The panelists and audience polling painted a remarkably consistent picture of the current state of R&D digitalisation.
Organisations are eager to embrace AI, but many still wrestle with fragmented data, disconnected workflows, legacy systems, and governance challenges. Before AI can deliver its full value, organisations must establish the digital foundations that make advanced analytics possible.
In the race toward AI-enabled R&D, success depends less on adopting the latest algorithms and more on solving the fundamental data and workflow challenges that still stand in the way.
