In the not-so-distant future, a virtual agent could buy your dog’s toys, settle your overseas invoices and flag potentially fraudulent charges on your credit card. All before your morning coffee is done brewing.

The exciting part? That future is arriving faster than anyone expected. And companies like Worldpay, one of the largest payment processors in the world, are preparing for it at full throttle.

“We’re quite bullish on agentic checkout and agentic commerce,” Nabil Manji, SVP head of FinTech growth and financial partnerships at Worldpay, told PYMNTS during a conversation for the June “What’s Next in Payments Series: Secret Agents.” 

The use of agentic artificial intelligence (AI), or autonomous AI agents that can reason, plan and transact on behalf of users, represents a seismic shift in commerce infrastructure. Welcome to the age of secret agents. Not in trench coats, but in code.

“Payments companies have been using machine learning and AI for years, if not decades,” Manji said. “One of the prerequisites for leveraging these tools is a large, rich dataset. And there’s a lot of data in payments and financial services.”

The challenge, of course, lies in making that data useful: clean, harmonized and actionable. But once that’s done, the payoff is exponential.

“Payments is a game of scale,” Manji said. “Minor improvements, whether it’s in fraud reduction, chargeback defense or authorization performance, can have a huge impact.”

We may not see it happening. In fact, we’re not supposed to. Like the best secret agents, AI is doing its work in the background, silently transforming how the world shops, pays and transacts.

And once the new system is in place, there’s no going back.

Rise of Machine-Led Transaction

Over the last 18 months, Worldpay has explored nearly 100 AI use cases, gradually homing in on a handful with the most strategic impact. Manji cites fraud and risk management as one of the most immediate external applications.

“We acquired a company called Ravelin a few months ago, which is at the intersection of AI and fraud prevention,” he said. “We’re putting our money where our mouth is.”

Read more: Mining the Insights Hidden in Your Customers’ Payment Inquiries

Another focus: enhancing customer service with generative AI. Rather than making customers wait while agents search databases, Worldpay is building tools that offer immediate response suggestions and surface relevant documents in real time.

“It gives our reps a starting point that allows them to respond much faster,” Manji said.

Internally, Worldpay is deploying AI to streamline developer workflows and customer onboarding.

“Like others in the tech space, we’re looking at tools like GitHub Copilot to help our devs write and test code faster,” Manji said. “And we’re actively using AI to improve our customer onboarding and underwriting journey — something the whole financial services sector has been trying to crack for years.”

AI’s value, ultimately, lies in optimizing processes that are both data-rich and time-consuming, a combination that financial institutions are all too familiar with. And if today’s AI-driven shopping feels like a supercharged personal assistant, it’s still only scratching the surface of what’s possible. 

But what happens after the selection is made illustrates the growing pains in agentic commerce. Behind the scenes, it’s a Frankensteinian patchwork. Most AI agents today execute transactions by filling out guest checkout forms — hardly the gold standard of conversion or security. That opacity creates headaches not just for customer service, but for fraud prevention, compliance and risk management.

“The front end is amazing,” Manji said. “But on the back end, it’s like we’ve gone backwards.”

Whose Fault Is It Anyway?

As agentic commerce grows, so do liability concerns. If your AI assistant buys the wrong T-shirt, who’s responsible? The consumer? The merchant? The AI platform? That ambiguity extends to chargebacks, fraud, and dispute resolution — critical areas that demand industrywide standards.

“There’s all these questions around where does risk sit, where does liability sit,” Manji said. “That needs to be figured out before agentic AI scales meaningfully.”

“We need things like identity verification and attestation,” he added. “What are the tokenization standards? What are the fraud implications? The network implications?”

Until those questions are answered, AI agents may be powerful but could be ultimately limited for consumers.

At the same time, as financial institutions race to deploy AI, so, too, do the fraudsters. The result? A cat-and-mouse game where both sides deploy algorithms in a digital arms race.

“We already know they’re using AI to make their attacks more sophisticated and learn faster than humans can,” said Manji, who believes AI-driven fraud detection will shift from rule-based systems to AI-augmented decision-making.

“Today, when a transaction trips a rule, someone reviews it manually,” he said. “But with AI, the system can say, ‘Here’s what triggered it. Here’s our perspective. We recommend approval or denial.’”

“It’s an augmented decision, not a binary one,” he added. “And that’s where I see most of this going.”

For Manji, the consumer experience is where AI agents will make their most visible, lasting impact. “It took a decade for people to get used to buying online. Then another decade for mobile,” he reflected. “It might take several years for consumers to start using agent-driven shopping. But once they do, they’re never going back.”

Consumers don’t unlearn convenience. And as agentic AI reduces shopping time from minutes to seconds, and potentially removes the need for interaction altogether, the old model of search, then scroll, then pay will soon feel as antiquated as dial-up internet.

“Merchants, acquirers, issuers, networks; we all need to think about how we’re preparing for that future,” said Manji. “Because once consumers adopt agent checkout, it’ll become the predominant form of lead generation and purchase.”