Merck and Mastercard Are Seeing Real Agentic AI Results. Both Say the Plumbing Came First.

Why It Matters

By embedding AI agents in core processes, Merck shortens time‑to‑patient for new therapies and slashes marketing overhead, setting a template for regulated industries. Mastercard's experiment shows the technology’s potential to automate complex, high‑risk financial operations without compromising trust.

Key Takeaways

• Merck's AI agents cut drug discovery cycles by one third.
• Marketing drafts achieve 99% compliance, shipping up to 80% faster.
• Infrastructure spans 2,500 AWS accounts, Azure, GCP, 47 edge sites.
• Guardrails use AI‑to‑AI checks, boosting confidence and reducing hallucinations.
• Mastercard pilots agents for dispute workflows, aiming to cut costs, preserve trust.

Pulse Analysis

Enterprises that rush into agentic AI without solid foundations risk creating fragmented, hard‑to‑manage systems. Merck’s experience illustrates why a "plumbing‑first" approach—building a unified, multi‑cloud backbone, standardizing data ingestion, and establishing secure agent registries—pays off. With 2,500 AWS accounts, Azure and GCP ties, and dozens of edge locations, the pharma giant can route agents to the right datasets, whether in Oracle, SQL, or unstructured repositories, ensuring each AI interaction has the context it needs to act responsibly.

The payoff is tangible. In research, AI‑driven agents trimmed a discovery cycle by roughly 33%, effectively shaving a year off the path to market for potential therapies. On the commercial side, AI‑generated marketing drafts now hit 99% compliance, collapsing review loops from months to days and accelerating rollout by up to 80%. Merck mitigates hallucinations by layering AI‑to‑AI validation, where one model’s output is vetted by another, raising confidence scores before human governance steps in. This guardrail strategy balances speed with regulatory rigor, a critical mix for heavily supervised sectors.

Mastercard’s parallel effort shows the model’s versatility beyond pharma. By deploying agents to orchestrate chargeback and fraud dispute workflows—tasks that blend structured transaction data with unstructured consumer narratives—the payments network aims to cut operational costs while preserving consumer trust. The firm emphasizes risk assessment from day one, weighing acceptable error rates against reputational stakes. Together, these case studies signal that when organizations invest in robust infrastructure first, agentic AI can become a scalable, low‑risk catalyst for efficiency across regulated industries.