Case in point: Mastercard announced Jan. 27 the Mastercard Agent Suite, a set of services and customizable AI agents that it said will help businesses build, test and deploy agents in day-to-day operations.
Kaushik Gopal, executive vice president of business and market insights at Mastercard, said leaders should treat agentic AI as a shift in how work gets done.
“Ultimately, when we get to this agentic world where a lot of activity is delegated, it is ensuring that there is a highly secure framework that is protecting that activity,” Gopal told PYMNTS. “That’s what creates a high level of trust and confidence from ecosystem participants to engage in that space.”
Mastercard is positioning Agent Suite as an answer to a practical problem. Many companies want to move fast, but many lack the foundation to use AI agents safely at scale. One-third of enterprise software applications are expected to incorporate agentic AI by 2028, Mastercard said in a press release, citing data from eMarketer, and the payments giant said it “expects a significant percentage of customer interactions and operational tasks to be supported by AI agents by 2030.”
That is where readiness comes in.
Advertisement: Scroll to Continue
“Readiness is the new competitive advantage,” Gopal said, adding that preparation helps companies turn innovation into outcomes faster.
Readiness is a set of building blocks, he said. Leaders need an AI-ready culture. Teams need training so business and technology groups can make decisions using shared language. Data needs to be structured, meaning organized and labeled, so software can use it reliably. Companies also need clear rules and ownership, plus the security controls that keep agents within bounds.
Agent Suite combines technical support with customizable agents, built on Mastercard’s payments expertise, data-driven insights and technology platforms, according to the press release. It also leans on a global advisory team of 4,000 advisors. The platform is expected to be available in the second quarter of this year.
What Enterprises Are Struggling With Most
Asked what is hardest for enterprises right now, Gopal said companies are at different stages. Most would not call themselves fully “AI native,” meaning they are still building the skills, processes and data foundations needed to use AI as a core capability.
“There’s a lot of upskilling, and there’s a lot of investment that’s required,” he said. “Then there’s policy of how I would deploy. Safety, security and trust have to underpin all of that.”
For many organizations, the bottleneck is often operating decisions, such as who approves an agent, what data it can see, how results are checked and when a person must step in.
Mastercard starts with business priorities and maps processes with a cross-section of customer leaders from business, operations and technology, Gopal said. Those sessions have accelerated since Mastercard’s Agent Pay announcement last year, as more clients move from curiosity to planning.
To prioritize automation, Gopal said he uses the impact-versus-difficulty test. Some high-value uses take longer because they require pulling data from many sources. Disputes are an example, since resolving a dispute can require receipts, shipping records and transaction data.
So, many roadmaps start where value is clear and data is already usable. Early wins build confidence and reveal what needs to be fixed before agents move into higher-risk work.
Buy, Build or Partner
Gopal said banks and merchants are asking a familiar question as AI agents move from demos to real work. Should they buy, build or partner?
He said he expects a mix. Some firms will build agents in-house and run them on their own systems. Others will license an agent or partner with a provider when that provider brings assets that are hard to replicate, such as broad data signals, hands-on expertise or platforms that can improve results.
“It will be build and deploy for myself, license and partner with someone, or completely outsource,” he said.
He pointed to fraud tools as an example. Mastercard’s fraud scores can add value because Mastercard can see patterns across markets, and that can complement a bank’s own data. He said he also expects the landscape to get more complex over time as agents start coordinating other agents.
Mastercard is pitching Agent Suite as the “partner” option, pairing customizable agents with technical support and its global advisory team to help customers move from plans to deployment.
Early Use Cases
The initial Agent Suite use cases focus on product discovery for banks and conversational shopping for merchants, Mastercard said in the press release.
In the bank scenario, an agent could recommend a product, such as a travel card or fee-saving account, and explain why it fits. Banks can test offer scenarios, trigger personalized campaigns and track performance to improve outcomes and drive portfolio growth.
In the merchant scenario, merchants can configure rules for inventory, margins, promotions and brand voice. An agent can then guide a shopper in a conversational way across channels, offering real-time recommendations based on preferences while following the merchant’s rules.
These entry points reflect what Mastercard is hearing across “hundreds of interactions” with banks and merchants, Gopal said. They also fit the impact-versus-difficulty test, since the starting data is often more structured, and the customer benefit is easy to see.
The promise of agents is delegated action. The risk is delegated action without control. Privacy, consent and intent become even more important when software is acting on a person’s behalf, Gopal said.
In delegated commerce, intent means capturing what the customer asked the agent to do. Consent means the agent should only use customer data when the customer has authorized it. Identity and authentication matter because the human is not always present at the moment an agent acts.
Agents built through Agent Suite will use privacy and responsible AI “by design” and will follow the company’s security principles, the press release said.
Gopal said he will know agentic AI has reached critical mass when adoption shows up in real activity, not just interest. He compared the shift to tokenization, which replaces sensitive card numbers with a secure digital token, and he pointed to Mastercard’s public goal of moving to a fully tokenized payments ecosystem by 2030.
For agentic AI at Mastercard, the signal is the share of customer interactions and operational tasks supported by agents and how quickly that share grows as enterprises gain confidence in the guardrails.
For all PYMNTS digital transformation, AI and B2B coverage, subscribe to the daily Digital Transformation, AI and B2B Newsletters.
