In the past quarter century, the business of online shopping has completely changed the global economy. From barely a blip at the start of the millennium, e-commerce now accounts for roughly 16% of all U.S. purchases, and that number is only going up from here.

Along the way, retailers and payment processors have had to stay ahead of the curve, deploying ever more advanced methods to verify and secure online transactions. Now we’re on the cusp of another major tech-driven shift.

If you’ve been following the conversations around artificial intelligence, then you know it’s set to revolutionize everything from software development to movie production. Today, three in ten teens use AI chatbots every day, whether by typing in a website or talking to a smart speaker. When those chatbots are enabled to take actions, such as sending emails, ordering food and the like, they are called agents.

More recently, people are using these agents to manage their purchasing needs, whether it’s groceries, apparel or major appliances. And this next wave of online shopping has a new name: agentic commerce, or a-commerce.

At a minimum, a-commerce could mean not only skipping the painful research portion of online shopping, when you spend minutes or hours or days reading reviews and comparing prices. It could also include ordering the agent to execute your purchase — a process often bogged down by verifying payment information, using promo codes or reentering the proper shipping address. Instead you could simply say, “Buy me the best pair of blue pants in my size for under $50 shipped,” and be done with it.

As the ability to complete shopping tasks by proxy becomes a reality, it’s also beginning to pose a few key problems. The first is verifying what the customer actually wants. When we humans do the comparison shopping and clicking and tapping ourselves, we do a pretty good job of ordering the right thing. Sure, the inseam on those pants might be a little short, or maybe the color is darker than it appeared in pictures, but there’s a good chance you’ll order something close to your needs.

When it comes to a-commerce, there’s a real possibility your designated agent might make a mistake and order you a pair of trousers in teal. Or maybe it decides to go with the British definition of “pants” and orders you a pair of cerulean underwear. These are understandable mistakes, but ones that you would probably never make yourself. So how can we ensure these sorts of misread intentions don’t happen?

Secondly, there are incredibly complex systems already in place designed to quickly identify and block fraudulent transactions, searching for deviations like purchases made in the middle of the night or from a location hundreds or even thousands of miles away from the customer’s home. This could make a-commerce transactions appear suspicious, as an agent might transact at odd hours or across geographies, or perform rapid, repeated purchases in ways that resemble the behavior of a fraud bot, resulting in purchase declines and undermining consumer trust. How can those same anti-fraud systems evolve for a future when legitimate transactions are being made by automated systems?

And finally, when mistakes do occur, who is at fault, and who is liable for making it right? Should the retailer have to cover the cost of returning those blue briefs? Should whoever developed your AI agent be responsible for rectifying the error? Or is it still your fault if you prompted but delegated the purchase?

That’s what we’re answering right now by building the standards for a-commerce with a coalition of partners, including the AI platforms themselves. Last year, we introduced Mastercard Agent Pay, which instills security and transparency in payments made by an AI agent before, during and after they’re made. At the center of the program are Mastercard Agentic Tokens, which harness the tokenization technology that already supports and secures payments worldwide. Tokenization enables AI agents to be uniquely tied to individual users, safeguarding payment credentials and allowing transactions to move forward seamlessly without constant user involvement.

But there’s more work to do in establishing protocols and rules that can scale to millions of merchants of all sizes, making a-commerce transparent and seamless on both sides of the  transaction. That includes ensuring that people maintain complete control over what the agent can purchase on their behalf; enabling every player in a transaction to recognize when an AI agent is at work; and even using AI agents to strengthen authentication to prevent bad actors from infiltrating the ecosystem.

Whatever other challenges lie ahead, our goal is to ensure that secure payments are a fundamental and trusted part of a-commerce throughout the entire life cycle of a purchase  — from the moment of need to the satisfaction of receipt.