The overall architecture of pharmacovigilance (PV) has changed very little over the last couple of decades. As AI-driven process automation advances from theoretical possibility to deployable reality, there is a growing case for looking beyond incremental performance gains and reconsidering the discipline from the ground up given that a fundamentally different approach is within reach.
From reactive to real-time: the regulatory direction of travel
The gap between regulator intent and operational reality around PV is now quite marked. The European Medicines Agency (EMA) articulated several years ago that, by 2030, pharmacovigilance (PV) for key new medicines should support real-time regulatory decision-making — taking the discipline from reactive adverse event collection towards proactive, continuous safety monitoring.1 Good Pharmacovigilance Practice (GVP) Module IX on signal management reinforces this expectation, positioning early detection and prompt evaluation as central objectives.2 Across the Atlantic, the FDA’s Sentinel Initiative is targeting comparable goals.3
If safety cases could be processed automatically — extracted, coded, medically reviewed and entered into a safety database within hours of receipt — signal detection teams would no longer be working from periodic snapshots. They would have access to real-time, structured data as a matter of course. Routine monthly listing cycles would become redundant. The seven-day Suspected Unexpected Serious Adverse Reaction (SUSAR) reporting timeline4, calibrated around manual workflows when first established, would begin to look more like an anachronism than a performance standard. With near-instantaneous processing, reporting within one or two days would be possible, with corresponding improvements in the speed of patient protection.
More data = more noise
Greater data availability does not automatically translate into sharper signal intelligence and, to a large extent, the proliferation of data sources over recent years has made the distillation of meaningful safety insights harder rather than easier. Sequential, manual workflows were not designed to cross-analyse such a spectrum of sources rapidly or coherently, while siloed and fragmented data makes it harder to understand. In reality, a multiplication of data means a rise in noise, and greater difficulty in finding what matters.
The duplication problem in adverse event reports is significant in its own right. Research conducted across seven major pharmaceutical companies by TransCelerate BioPharma identified a mean of three submissions per case version across 2.5 million case versions, with a meaningful proportion reaching ten or more health authority recipients.5 At the same time, external factors — heightened media coverage of a particular drug class, or public concern about a newly authorised product — can generate surges in reporting volumes that do not correlate with any actual underlying product risk.
AI-powered signal evaluation offers a practicable route through this problem — not by replacing expert analysis, but by making the latter substantially more effective. Harnessing agentic AI technology transforms the ability to distinguish genuine statistical patterns from reporting artefacts, surface the cases most likely to represent true safety signals and direct expert attention accordingly, because of agents’ ability to observe, reason, and recommend. Once the agents are performing the data collection, analysis, and cross-triangulation of multiple data sources. The safety scientist can then focus on assessing the recommendations and making decisions, the safety scientist is free to focus on assessing the recommendations and making decisions.
Reconsidering the case for human review
To achieve near-instantaneous case processing and real-time data availability for signal management, it is important to address the rate-limiting step of human review. Concern about reducing human involvement in case processing is understandable, but this assumes that manual performance is consistent and reliable — an assumption that does not hold up well to examination. Routine inspection findings relating to Individual Case Safety Report (ICSR) quality exist because human-dependent workflows are difficult to standardise at scale. Quality control sampling at five or ten percent of case volume provides no guarantee of systematic quality, and the effects of fatigue and interpretive inconsistency become more pronounced as processing volumes rise.
An automated system with full-coverage quality monitoring — and a governance layer that detects model drift and enables systematic retraining — represents a substantive improvement on such scenarios. When quality problems arise, teams can more reliably instruct agentic agents to reference updated materials than redirect hundreds of case processors across multiple sites. AI-powered processing is also inherently more auditable: every decision can be logged and every deviation can be rendered traceable in a way that manual processes cannot replicate.
The industry has navigated comparable transitions before. When PV functions began outsourcing case processing to contract research organisations in the late 1990s and early 2000s, the concerns raised at that time closely mirror those being voiced today about AI: how would sponsors maintain oversight? How would quality be assured at a distance? Those questions were resolved, and that transition is now so well established that the concerns barely feature.
The goal being described here is not autonomous pharmacovigilance, in any case. It is, rather, augmented pharmacovigilance: AI managing the high-volume, labour-intensive tasks to which it is well suited; human expertise being reserved for interpretation, regulatory dialogue, benefit-risk assessment and communication with healthcare professionals.
Realistic timelines for this scale of change
The technology building blocks to enable this step-change in PV workflow is now in active development. Automated ICSR processing pipelines, agentic AI coding assistants, continuous quality review layers, cross-domain data integration platforms and AI-powered signal evaluation tools together represent a substantial and increasingly deployable capability set. Layered on top of this, there must also be organisational readiness, however — the willingness to deploy these tools at scale, underpinned by governance frameworks robust enough to withstand regulatory scrutiny of the outputs.
The case for change is particularly acute for products approved after exposure in relatively small patient populations. Products reaching the market having been studied in only a few hundred patients — increasingly common in areas of unmet medical need — present an especially demanding signal management challenge. The safety database at launch may be limited, making early post-authorisation signal detection correspondingly more critical; faster detection means earlier intervention.
Organisations that get ahead with this transition could, within five years, be operating with near-instantaneous case processing as standard. EMA’s ambition for real-time PV by 2030 will not be delivered by process optimisation alone; it will also require infrastructure investment and organisational commitment. For those prepared to make the leap, the potential benefits extend well beyond operational efficiency, to include earlier detection, stronger patient protection and a safety function whose human expertise is directed to where it can have the greatest impact.
About the author
Lucinda Smith is ArisGlobal‘s chief safety product officer. She previously spent more than two decades working in frontline scientific and strategic pharmacovigilance and drug safety roles at a major pharma brand.
References
1 How will pharmacovigilance look in 2030? European Medicines Agency, March 2020. https://www.ema.europa.eu/en/news/how-will-pharmacovigilance-look-2030
2 Guideline on Good Pharmacovigilance Practices (GVP) Module IX — Signal Management (Rev. 1). European Medicines Agency, November 2017. https://www.ema.europa.eu/en/human-regulatory-overview/post-authorisation/pharmacovigilance-post-authorisation/signal-management
3 Leveraging real-world data for safety signal detection and risk management in pre- and post-market settings. Frontiers in Drug Safety and Regulation, September 2025. https://www.frontiersin.org/journals/drug-safety-and-regulation/articles/10.3389/fdsfr.2025.1626822/full
4 ICH Guideline E2A: Clinical Safety Data Management — Definitions and Standards for Expedited Reporting. International Council for Harmonisation; in force since 1994. https://www.ich.org/page/safety-guidelines
5 Individual Case Safety Report Replication: An Analysis of Case Reporting Transmission Networks. Drug Safety, January 2023. https://pmc.ncbi.nlm.nih.gov/articles/PMC9870831/
6 Might We Come Together on a Paradigm Shift to Manage ICSRs with a Decentralised Data Model? Drug Safety, April 2025. https://doi.org/10.1007/s40264-025-01539-4
