Agentic AI Results at Merck and Mastercard: Infrastructure as a Foundation

Merck and Mastercard leverage agentic AI to enhance efficiency across various sectors. Their success hinges on robust infrastructure, transforming processes from drug discovery to financial dispute resolution.

Recent advancements in artificial intelligence at Merck and Mastercard illustrate a significant transformation in operational processes across pharmaceutical and financial sectors. The utilization of agentic AI at these organizations is unlocking substantial efficiencies, yet the pivotal element behind these gains is robust digital infrastructure.

At Merck, the pharmaceutical giant has harnessed AI agents to streamline drug discovery cycles by a third and expedite the delivery of compliant marketing materials by up to 80%. Merck’s VP of Digital Platforms, Sean Finnerty, credits these advancements not solely to the AI itself but to a foundational infrastructure that was meticulously established first.

Infrastructure as a Catalyst

The emphasis on building a solid ‘plumbing’—a term used by Finnerty to describe the digital foundation—draws lessons from the early cloud computing era. Merck’s adoption of cloud infrastructure supports an extensive network of 2,500 AWS accounts alongside numerous Microsoft Azure and Google Cloud Platform integrations. This robust setup facilitates the seamless deployment and management of thousands of AI agents.

Finnerty highlights the complexity of registering, securing, and integrating these agents with appropriate tools and data, a task made feasible only with a well-orchestrated infrastructure. This setup also involves managing ‘many, many petabytes’ of data across various platforms, showcasing the necessity of comprehensive scaffolding to deliver meaningful context for AI operations.

“The challenge is to make these processes easy and frictionless while ensuring security and integration with model context protocols and agent-to-agent communications,” Finnerty explained. This approach ensures that workload distribution across platforms like AWS and GCP is seamless.

Enhanced Drug Discovery and Compliance

The integration of AI has notably accelerated Merck’s drug discovery and compliance processes, reducing discovery cycles by up to a year and enhancing market readiness. This allows Merck to deliver life-saving therapies to patients significantly faster. The shift from human-in-the-loop processes to AI-driven compliance checks—in which AI delivers drafts that are ‘99% right’—has revolutionized the review cycle from months to mere days.

Moreover, app modernization is no longer a laborious, expensive endeavor. AI agents aid in code refactoring, API interaction documentation, and authentication checks, previously requiring substantial expenditure and time.

Mastercard’s Financial Services Shift

Similarly, at Mastercard, agentic AI is being strategically applied to optimize transaction and dispute workflows, areas traditionally labor-intensive and prone to inefficiencies. Chief Data Officer Andrew Reiskind notes the intricacy of these workflows, which involve structured and unstructured data elements and require precise decision-making.

AI agents are deployed to handle complex tasks in these workflows, balancing deterministic and probabilistic decision-making. This shift not only promises efficiency but is also expected to maintain consumer trust—a critical factor in financial services.

Complexities and Mitigations

While the benefits are clear, challenges such as AI ‘hallucinations’—where AI generates scenarios detached from actual contexts—persist. Merck tackles this by engineering guardrails and implementing AI checks to validate outputs, thus minimizing errors.

At both companies, the journey involves assessing and accepting certain operational risks. Reiskind likens this to serving a wrong sandwich order versus a life-threatening allergen, underscoring the importance of differentiating and mitigating risks in AI deployment.

Conclusion

The emergence of agentic AI at Merck and Mastercard exemplifies the transformation driven by foundational infrastructure. These cases underscore that meaningful AI deployment extends beyond the technology itself, requiring a comprehensive, secure, and scalable infrastructure.

The observed shift promises continued advances in efficiency, heralding a new era of automation in pharmaceuticals and financial services. As these systems evolve, monitoring the integration and adaptation within human workflows is vital.

Observation recorded.