Merck and Mastercard are seeing real agentic AI results. Both say the plumbing came first.

Merck says its AI agents are already producing measurable results, including cutting one drug discovery cycle by 33% and speeding compliant marketing material delivery by 70% to 80%. VP Sean Finnerty says those gains depend on building the underlying infrastructure first, rather than deploying disconnected one-off agents that become technical debt.

The article underscores a broader enterprise AI lesson: useful agentic AI depends less on flashy demos than on identity, security, data access, and orchestration layers that can scale. Companies that standardize this plumbing early may move faster as agent deployments multiply across clouds, edge locations, and fragmented data systems.