No one wants to experience buyer’s remorse. That’s why before you make large purchases, you probably do a bit of research.

If you’re using artificial intelligence to do that research, asking ChatGPT which machine makes the best espresso, for example, you’ve already experienced the first wave of AI commerce. Indeed, millions of consumers now use AI to research purchases before they ever visit a retailer’s website.

And the next wave of AI commerce is going to be even bigger: soon, AI agents will go beyond recommending products and do the shopping for us.

The technology already exists, but something essential is holding it back: trust.

As a technology and innovation leader focused on digital identity, trust, and fraud prevention, it’s a conversation I’m hearing over and over again among retailers, payment providers, financial institutions, and tech companies. There’s real excitement around agentic commerce, but there are questions as well:

  • How do we know an AI agent is acting on behalf of a real customer?
  • What happens if something goes wrong?
  • And how do retailers embrace this new way of shopping without introducing risk?

Trusting AI’s Advice is One Thing. Trusting its Decisions is Another

Consumer-facing AI hasn’t been around very long, yet consumers are already remarkably comfortable using it as a shopping assistant. Generative AI tools have become indispensable research partners, in part because the consumer is still making the final purchasing decision.

However, there’s a big difference between asking AI, “What should I buy?” and saying, “Here’s my credit card. Go buy something for me.”

Crossing that chasm requires a leap of faith, and it’s understandable why consumers might hesitate at the edge. After all, even traditional e-commerce sometimes goes wrong, from delayed deliveries and damaged products to fraud and payment disputes.

But while the commerce ecosystem has established systems to manage those situations, and consumers generally know what to expect when something goes wrong, agentic commerce has yet to build those same guardrails.

Retailers have their own reasons to be wary. The industry has invested heavily in fraud prevention, authentication, and payment security to reduce risk when consumers make online purchases, but what happens when the person behind a transaction isn’t a person at all? How does a retailer know who authorized the agent, what it’s allowed to do, and who is responsible if the purchase isn’t what the customer intended?

In fact, AI is fueling both innovation and fraud simultaneously. Generative AI helps bad actors make scams more convincing, automate attacks, and lower the barriers for attacks at scale. Retailers will need to evolve their identity and fraud protections just as quickly.

And in retailers’ eyes, all bots used to be the bad guys, scraping websites, overloading systems, manipulating prices, and fueling fraud. Retailers became very good at identifying and blocking automated traffic. But agentic commerce has turned this dynamic on its head: now bots may carry out innocuous tasks such as comparing prices or waiting to purchase an item the moment it comes back in stock.

Yesterday’s bad actor may look just like tomorrow’s best customer, and retailers will increasingly need ways to distinguish between the two.

Trust is Everyone’s Job

One thing gives me confidence about these challenges: retailers, payment providers, and technology companies are working to solve them together.

Commerce has always depended on collaboration. Every online purchase already relies on merchants, payment providers, financial institutions, cybersecurity companies, fraud teams, and identity technologies working in concert behind the scenes.

With agentic commerce, the industry has the opportunity to build on the infrastructure that’s already in place. Existing payment networks, fraud controls, identity verification, and merchant risk systems provide a strong foundation for supporting AI-driven transactions.

This is good news for consumer trust as well. Consumers don’t need to learn an entirely new commerce ecosystem; they only need the confidence that the one they’re already using can safely support AI agents acting on their behalf.

Trust will also benefit from smaller, simpler breakthroughs. When people imagine AI shopping agents, they often picture handing over every purchasing decision to an autonomous assistant. However, it’s unlikely that this will become a reality overnight. More likely, agentic commerce will begin with straightforward, well-defined tasks where the consumer sets the rules and the AI does the work.

For example, rather than assigning an AI agent to manage their entire holiday shopping list, a consumer might feed it a prompt with clear boundaries, such as, “If this item comes back in stock for less than $100 and can arrive before Dec. 15, buy it.”

Not only does that sound more appealing than refreshing a retailer’s website every few hours, it also sounds relatively foolproof. And every successful interaction will help consumers become a little more comfortable with what AI can do.

Agentic Commerce Depends on Trust

In commerce, innovations come first and trust follows after.

More than a century ago, consumers who first ordered from the SEARS catalog had to trust that their order forms and money orders would arrive safely through the mail. Twenty years ago, consumers first using mobile pay had to trust that these transactions were secure and no one could steal their information.

Over time, across millions of safe and successful transactions, each of these technologies became mainstream.

The AI age will be no different. The technology for AI-powered shopping is here, but it’s up to retailers, payment providers, and technology companies to build the trust into the system that will allow it to scale. Once consumers feel confident, AI commerce will flourish.

Kathleen Peters is chief innovation officer at Experian, a global data and technology company powering opportunities for people and businesses around the world.