In the dawning era of Artificial Intelligence, hype often obscures practical reality. However, two giants from entirely different sectors—pharmaceutical leader Merck and financial services titan Mastercard—are grounding the conversation by revealing the secret behind the successful adoption of "Agentic AI." Their message is clear: before AI agents can begin making decisions and executing tasks, a business must have its "plumbing" in order—meaning its data infrastructure and governance frameworks.
Merck and the Revolution in Drug Discovery
For Merck, the use of AI agents is not merely an experiment but a tool already delivering measurable results. According to Sean Finnerty, Vice President of Digital Platforms at Merck, the company has managed to cut drug discovery cycles by a third. In an industry where developing a new drug can take a decade and cost billions, a 33% reduction in time is a monumental achievement.
Furthermore, the company utilizes AI agents to create and review marketing materials. The compliance process for the industry's strict regulations, which previously required weeks of manual checks, is now accelerated by 80%. These agents don't just suggest copy; they verify legal validity, the consistency of scientific references, and compliance with FDA standards, freeing humans from grueling bureaucracy.
Mastercard: Security and Internal Efficiency
On the other hand, Mastercard focuses on leveraging Agentic AI to enhance transaction security and automate internal workflows. For a company handling trillions of dollars globally, the ability of AI agents to act autonomously to detect anomalies in real-time is vital. However, Mastercard emphasizes that this technology would be useless without a unified data system that allows the AI to access the enterprise's "truth" without delays.
The Lesson of the 'Plumbing'
The common reference point for both companies is the emphasis on infrastructure. The term "plumbing" refers to a company's ability to clean, organize, and connect its data so that AI can understand it. Without this groundwork, AI agents are like sophisticated engines connected to rusty pipes: the output will be either incorrect or dangerous.
- Data Governance: Who has access to what? Is the data valid?
- Interoperability: Can different systems "talk" to each other?
- Security: How do we ensure an autonomous agent won't breach security protocols?
Merck spent years building a digital platform that unifies data from labs to sales. This investment is what allows AI agents to function effectively today. As Finnerty points out, many organizations try to jump straight into AI without solving their data problem, which inevitably leads to failure.
The Transition from Chat to Action
The fundamental difference between the Generative AI we met in 2023 (like ChatGPT) and the Agentic AI of 2026 is the capacity for action. While early models could only write or analyze, agents can execute. They can open a support ticket, close a deal, order chemicals for an experiment, or block a suspicious card.
"AI is no longer a conversationalist; it is a partner that executes tasks. But a partner is only as good as the information you provide," says a market executive.
Challenges and the Future
Despite the optimism, challenges remain. Trust is the primary concern. How much autonomy should we grant an agent that can affect human health or the global economy? Merck and Mastercard follow a "human-in-the-loop" approach for critical decisions, using AI to prepare the ground and humans to give the final approval.
The example of these two companies points the way for other businesses. Artificial intelligence is not a magic wand but a structure. And like any structure, its strength depends on its foundation. For those who invested early in their data "plumbing," the era of agentic AI promises massive competitive advantages. For the rest, the road will be full of leaks and failures.
