By Émile Malan, Managing Director, Vicore Health
Moving Beyond “Human-in-the-Loop”
“Human-in-the-loop” has become one of the most common phrases in conversations about AI in pharma. Regulators use it. Vendors use it. Compliance teams use it. But in pharmacovigilance and regulatory affairs, it is the wrong way to think about AI.
A loop suggests oversight at the end of a process. The system runs, and a person checks the output. That model may work in low-risk environments. It does not work in regulated healthcare systems where accountability, judgement, and patient safety cannot be delegated to technology.
In pharmacovigilance and regulatory affairs, humans are not simply reviewing outputs. They are leading the process from beginning to end. They define the framework, apply the judgement, and carry the accountability. That is what human-in-the-lead actually means.
AI Scales Existing Systems - Good or Bad
This distinction matters because AI does not create wisdom or expertise on its own. It scales whatever already exists inside an organisation. Strong systems become more efficient. Weak systems become faster at producing weak outputs.
A pharmacovigilance programme with inconsistent signal-detection logic will not suddenly become mature because AI has been added. A regulatory team without clear processes or local market understanding will not become more accurate because an AI tool can generate polished responses in seconds. Technology amplifies the quality of the underlying system. It does not replace it.
This is particularly important in African regulatory environments, where teams are often operating under significant capacity constraints while managing increasingly complex regulatory and safety obligations.
The Opportunity and Risk of AI in Pharmacovigilance
Across pharmacovigilance functions, AI is already changing how work gets done. Literature surveillance, case triage, signal detection, and narrative drafting can now be performed far more efficiently than before. Tasks that once took weeks can often be completed in days or hours.
The productivity gains are real, especially in regions where adverse drug reactions remain significantly under-reported. AI offers an opportunity to strengthen pharmacovigilance systems at a scale that would be difficult to achieve through manual processes alone.
But speed is not the same as maturity.
If the underlying workflows, escalation pathways, or governance structures are weak, AI simply accelerates those weaknesses. Faster processing does not automatically produce better decisions. In some cases, it can increase risk by creating confidence in outputs that have not been properly interrogated.
Why Regulatory Affairs Faces the Same Challenge
The same challenge exists in regulatory affairs. AI can now draft dossier sections, summarise guidance documents, and support regulatory intelligence gathering. The outputs often appear highly coherent. The critical question is whether they are accurate, contextually appropriate, and regulatorily defensible.
Generative AI systems do not reliably communicate uncertainty. They can present weak conclusions with the same confidence as strong ones. In regulated environments, that matters. A single inaccurate statement in a submission can trigger delays, requests for clarification, or compliance findings.
This is why governance matters as much as technology.
Globally, regulators are beginning to focus less on whether AI is being used and more on how organisations control and oversee it. The emphasis is increasingly on accountability, validation, traceability, and lifecycle management of AI-supported systems.
That shift aligns closely with the idea of human-in-the-lead. The real question is not who reviews the final output. It is who owns the system, understands its limitations, and takes responsibility for the decisions being made.
Africa’s Opportunity to Build Differently
Africa has a unique opportunity in this space.
Many regulatory systems across the continent are still evolving digitally. While this creates challenges, it also creates the opportunity to build modern, AI-augmented operating models from the outset rather than retrofitting them later.
This matters because the scale of the challenge is significant. Regulatory authorities and pharmacovigilance teams across Africa are expected to manage growing product portfolios, expanding safety obligations, and increasing regulatory complexity with limited resources and uneven infrastructure.
Traditional capacity-building alone is unlikely to close the gap quickly enough.
Human-in-the-lead models supported by validated AI can help address this challenge more sustainably. The goal is not to replace expertise. It is to allow highly skilled regulatory and pharmacovigilance professionals to spend less time on repetitive, high-volume administrative work and more time on interpretation, strategy, and decision-making.
What Human-in-the-Lead Requires in Practice
In practice, human-in-the-lead requires three things:
- Strong governance frameworks before AI enters the workflow. Decision-making principles, escalation pathways, and compliance expectations must already be clearly defined.
- AI is used as augmentation, not replacement. Automation should support drafting, summarisation, classification, and pattern recognition while humans retain ownership of all consequential decisions.
- Structured regulatory intelligence and safety systems that allow country-specific requirements, local obligations, and historical context to be managed in a controlled and traceable way.
The Future of AI in African Healthcare Regulation
At Vicore Health, this is the model we believe will define the future of regulatory affairs and pharmacovigilance in Africa. Humans remain firmly in the lead, supported by validated, purpose-built technology that strengthens efficiency, visibility, and compliance across increasingly complex environments.
The future of AI in healthcare regulation is not about removing humans from the process. It is about enabling human expertise to operate more effectively at scale.
Automating large portions of the screening process does not remove the need for human expertise. It reallocates it.
By reducing manual workload, pharmacovigilance professionals can focus on interpretation, decision-making, and patient safety outcomes rather than repetitive tasks.
