As pharmaceutical companies explore AI for safety monitoring, submissions and labelling, a familiar challenge is emerging: automation can only go as far as the data allows.
Across the pharmaceutical sector, interest in artificial intelligence continues to grow. Companies are testing AI tools for regulatory submissions, safety signal detection, labelling harmonisation and other data-heavy processes. But while the technology is advancing quickly, many organisations are still being held back by a more basic issue: fragmented, inconsistent and poorly governed data.
Writing in DDW, Ian Crone of ArisGlobal argued that European pharmaceutical companies should treat data reform not simply as a compliance requirement, but as a foundation for the next stage of automation. The point was echoed during a Life Science AI Exchange Podcast panel discussion, where industry experts warned that decades of uneven progress on medicinal product data are now limiting the ability to deploy AI at scale.
From Compliance Task to AI Foundation
At the centre of the issue are ISO’s Identification of Medicinal Products standards, known as IDMP. These standards are designed to create more consistent definitions and improve the exchange of product information across the medicines regulatory network and beyond.
The European Medicines Agency is introducing IDMP through a phased programme built around substance, product, organisation and referential master data. In practice, this should make medicinal product information easier to reuse across manufacturing, supply chain, regulatory, quality and safety functions.
For data science teams, the wider lesson is clear. AI systems depend on data that is structured, trusted and reusable. Without common definitions, strong governance and reliable data flows, even promising AI pilots can struggle to move into production or deliver meaningful long-term value.
The Problem with Fragmented Data
Panellists warned that many pharmaceutical companies have left the groundwork too late. Peter Brandstetter of Accenture said the industry should have begun modernising product data flows 10 to 15 years ago. Remco Munnik of Arcana Life Sciences Consulting argued that AI cannot make sense of information that lacks structure and governance.
This is not a challenge unique to pharma. Many sectors are discovering that access to AI tools is no longer the main barrier. The harder challenge is creating the data environment needed for those tools to work reliably, transparently and at scale.
In pharmaceutical settings, the risks are especially clear. If product information is trapped in documents, stored inconsistently or managed in separate systems, AI-enabled processes can quickly run into limits. Poor data foundations can lead to poor outputs, slow adoption and reduced confidence in the technology.
Building a Common Data Language
The EMA’s Product Management Service is being developed as a central part of the IDMP transition. It uses FHIR as the basis for its application programming interface, helping to support a more standardised approach to exchanging product data.
This matters because AI and automation depend on more than volume. They depend on context, consistency and interoperability. A common data language makes it easier for information to move between systems, teams and processes without losing meaning along the way.
That could support practical use cases such as automated labelling updates, safety change management, translation workflows and more joined-up regulatory planning. Munnik pointed to prototype scenarios where structured product data allowed approved safety changes to be pushed into labelling and translations automatically, reducing manual work while improving consistency.
A More Strategic Role for Regulatory Data
The panel also suggested that regulatory affairs teams could become a more strategic force within pharmaceutical companies. Frits Stulp of Implement Consulting Group noted that these teams hold some of the most regulator-validated product information in the business.
Used effectively, that information could support wider portfolio questions, identify trends and inform planning beyond narrow compliance activity. In other words, regulatory data could become a valuable strategic asset, not just a record-keeping requirement.
The Data Science Takeaway
For data science professionals, the story is a useful reminder that AI readiness is not just about models, platforms or experimentation. It is also about the less glamorous work of data structure, governance, standards and stewardship.
Pharmaceutical companies may be working within a highly regulated environment, but the underlying challenge is familiar across industries. AI pilots can generate interest, but scalable impact depends on the data foundations beneath them.
For organisations that have fallen behind, catching up is still possible. But the message from the panel was clear: companies need to move beyond seeing data reform as a compliance chore and start treating it as a long-term strategy for trusted automation and AI-enabled decision-making.
References
https://www.ddw-online.com/why-pharma-must-refocus-its-data-efforts-if-it-really-wants-to-exploit-ai-41823-202605/
https://www.ema.europa.eu/en/human-regulatory-overview/research-development/data-medicines-iso-idmp-standards-overview/substance-product-organisation-referential-spor-master-data