Mastercard’s new service signals the shift from AI that speaks to AI that acts

For a technology often accused of being all talk and no action, artificial intelligence is finally getting a job. On January 27th, Mastercard announced the launch of its Agent Suite, a set of tools designed to help banks and retailers build, test, and deploy “agentic” systems. These are not your standard chatbots that merely regurgitate text; they are autonomous software entities capable of planning and executing complex tasks—such as recommending a travel card to a jet-setter or managing a merchant’s inventory—with minimal human oversight.

The move signals a shift in the corporate zeitgeist. If 2025 was the year of “GenAI” fatigue, 2026 is becoming the year of the agent. While traditional AI predicts and generative AI creates, agentic AI decides and acts. It operates in a “perception-reasoning-action” loop, sensing its environment and iteratively refining its behavior to meet a goal. In the words of Kaushik Gopal, a head at Mastercard, “readiness is the new competitive advantage”.

Yet, for all the corporate enthusiasm, a familiar scent of over-exuberance hangs in the air. Some analysts warn of an AI bubble beginning to deflate. Just as the dot-com era’s “eyeballs” gave way to a demand for profits, investors are now asking when the billions spent on GPUs will actually show up on the bottom line. Thomas H. Davenport and Randy Bean, columnists for MIT Sloan Management Review, suggest that agents are headed straight for the “trough of disillusionment” in 2026. They note that university experiments show agents still make too many mistakes for “prime-time” business involving significant sums of money.

The obstacles to adoption are more mundane than a robot uprising. Most enterprise data is still “trapped” in unstructured formats like invoices and emails. Without intelligent document processing to clean this digital attic, an AI agent is as useless as a butler with no keys. Companies like TileDB are attempting to bridge this gap with multimodal databases that allow agents to “reason” over everything from genomic data to clinical records without the lag of traditional systems.

There is also the matter of who, exactly, is in charge. While 39% of large firms now have a Chief AI Officer, there is little consensus on where they should sit. Some report to technology heads, others to transformation leaders, and only 30% to the chief data officer—a structural mess that likely contributes to the struggle of delivering measurable value.

Despite these growing pains, the direction of travel is clear. Roughly a third of enterprise applications are expected to incorporate agentic AI by 2028. Leading banks like JPMorgan Chase and BBVA have already established “AI factories” to churn out these models at scale. Marc Benioff, the boss of Salesforce, has even mused that he might be the last CEO of his firm to manage only humans.

For now, the digital doers remain more like interns than executives: helpful, prone to hallucinations, and requiring constant supervision. But as Mastercard’s new suite suggests, the business of AI is no longer just about writing a better blog post; it is about getting the work done.