Automating Drug Safety Processes at GSK with AI and Automation

At GSK, delivering drug-safety reports to the right specialist often takes several hours. Reports arrive through various channels like emails, paper mail, faxes, and data loggers. Each report must be read, prioritized, and allocated appropriately. During this process, essential details may repeatedly be re-entered and reconciled across systems.

“It’s not a process problem, it’s a data problem,” commented Lisa-Ann Chiriseri, GSK’s director of drug safety and pharmacovigilance applications.

Chiriseri aims to transform how information flows through pharmacovigilance, prioritizing specialist decision-making. She discussed this initiative during the Seeking Value video series, with Cognizant’s Mohammad Haque and Newsweek’s Gabriel Snyder.

Cognizant collaborates with GSK to automate drug-safety workflows. Haque noted that improving individual tasks offers limited benefits if subsequent teams must still reconstruct data:

“The problem has not gone away, it has just moved.”

Both companies work on streamlining the entire safety report lifecycle, from initial arrival to regulatory reports. They aim to maintain data integrity while cutting down on manual preparation steps.

GSK holds well-established safety systems as reliable information sources, but current systems should not constrain every workflow. Automation decisions start by identifying specific issues and evaluating technological solutions.

Chiriseri mentioned that repetitive tasks, even those performed by trained specialists, could benefit from AI and conventional automation. GSK employs these tools to classify safety reports and prioritize them accurately. This structured data processing helps preserve necessary context for further evaluations.

Haque envisions specialized AI agents facilitating a seamless pharmacovigilance process. He proposed an operation model with AI covering seven interconnected areas, from receiving and processing reports to supporting inspection requirements.

This model suggests having up to 40 specialized agents managing tasks across the life cycle, working in parallel to minimize delays between teams.

While GSK and Cognizant have initiated smaller automation projects, their broader vision involves a unified agent architecture to reshape organizational information flow.

Human judgment remains essential due to regulatory obligations in drug safety. Distinct tasks fall into two categories: those governed by clear rules and those needing domain expertise. Automation handles predictable tasks, while AI assists with complex data preparation.

The proposed model includes oversight: agents may prepare data for professional review, or manage workflows with oversight from employees who set thresholds and monitor output.

“We’re not asking for blind trust from our regulators and quality and compliance teams,” Chiriseri stated.

GSK integrates traceability within its systems, enabling reconstruction of data handling and decision processes. This ensures new capabilities comply with regulations and safeguards are in place before system deployment.

The companies also revisit measurement of technology investments. Haque remarked that standard automation metrics fail to reflect the real benefits:

“We start measuring the outcomes. And the big one is time.”

Targets include reducing case processing time and report preparation by 50%, and decreasing the time from identifying to resolving safety signals by 40%. They also aim for over 90% consistency in quality control activities.

Chiriseri noted that GSK is moving from trial phases to applying these solutions in production, with positive early results reinforcing their strategy. A key success indicator is how well employees adapt to new technologies. GSK supports workers to learn and improve processes actively.

The strategy involves what is termed as citizen development, allowing non-engineers to contribute to creating applications, drawing on their understanding of inefficiencies. Chiriseri emphasized reusing information and technology rather than rebuilding, connecting existing data for new workflows.

Haque advised other companies under strong regulatory control to map their processes and data thoroughly before introducing AI agents:

“Let the agents do the predictable work, make every decision visible and keep people accountable by design.”