It can be hard to reckon with the speed of innovation from the past few years, most notably with the grand introduction of AI. We’ve moved at a pace that nobody was prepared for. Until just recently, the notion that an agent could execute complex tasks on one’s behalf was an unattainable dream. However, as we’ve seen this month with the launch of Cloudfare’s Kitsurf, the internet is rebuilding the consumer experience and a new normal is underway.
There has been an unsurprising stream of Silicon Valley-fueled excitement around the concept of AI agents taking over the mundane daily tasks of the average person, but that excitement has struggled to transcend the confines of San Francisco. Because everyday consumers care about practical functionality rather than underlying mechanics, a major communication gap has emerged, severely restricting these discussions to tech-centric circles. What is largely absent in these conversations, and has kept this topic from growing legs, is the cumulative effect that this change will have on the very foundations of the shopping experience and customer loyalty.
We’ve heard a lot of arguments for the practical usage of agents for the average consumer, but less has been said about the shift underway in how businesses build trust with the people who buy from them. That relationship has run on networks and human judgment for decades, and agentic commerce puts it at risk. An industry that has run on networks and relationships faces a breakdown of this very human system. Still, this trajectory isn’t set in stone.
Agents can and should be tied back to a real identity. That’s what makes it possible to keep optimizing for agent compatibility without giving up the relationship underneath it. They serve each other better than they operate singularly.
What gets lost when an agent shops on price alone
An agent trained to primarily consider price, speed, and efficiency isn’t equipped to replicate human trust and loyalty without access to purchase history. A customer’s history and loyalty are exactly the information an agent needs to be useful. An agent informed about the purchaser, as well as programmed to consider efficiency, is one that actually serves the consumer.
This behavior is not siloed either. The same pattern shows up in B2B procurement, where buyers increasingly let AI tools generate supplier shortlists before a human conversation happens. Over time, relying solely on these metrics risks eroding the long-term value that the human touch provides.
Society at large finds comfort in operating within the parameters of black and white, polarization, and extremes. When we hear about the pitfalls of AI, we have the instinctual reaction to swing to the fully-human operated end of the spectrum and vice versa. Instead of one or the other, what if we require a ‘yes and’ solution? Rather than fully eliminating the human element and allowing bots to operate with free rein, the efficient solution they bring to the table makes for a great collaboration rather than a human replacement.
The trust deficit in agentic commerce
Nothing about agentic commerce forces anonymity between merchant and customer. That default came from how quickly the first agentic commerce tools were built, not from any limit in what the technology can do.
We can see firsthand at companies like Google that when a platform mediates between business and customer at scale, it ends up owning the relationship. The business defaults to optimizing for the platform’s logic rather than prioritizing the customer’s trust. Agentic commerce risks the same outcome: merchants reduced to interchangeable line items an agent picks on price, with no path back to a direct relationship.
Visa’s research found that close to nine out of ten people surveyed want to see how an AI shopping agent reaches its decision, and roughly half of that group would stop using one if they lost control over how it operated. A strong data point that only further reinforces just how much users want verification as opposed to blind trust. Transparency must become a core feature rather than an afterthought.
Once upon a time, e-commerce faced a similar issue as it was introduced to the world. Trusting that your credit card information was safe, even on a secure website, did not come without trials and tribulations. Yet, they solved their trust problem through standards rather than accepting fraud as the cost of doing business online. Agentic commerce demands the same fix.
Building trust into agents, not around them
Consumers’ top three conditions for delegating a purchase to an agent are spending caps, instant revocation, and easy cancellation. Meanwhile, 75% of merchants say real-time permission revocation will be critical to adoption. Both sides are encountering the same issue.
It’s imperative that credentials stay with the person, qualified by the agent asking before it acts on anything sensitive. Every agent should carry a verifiable identity separate from any single session, closing the impersonation gap on both sides of a transaction. Permission should stay granular and revocable, allowing an agent to browse and cart without holding payment or account access, and identity should persist across sessions, so a retailer can attach loyalty and history to the person behind the agent, instead of treating every purchase as anonymous.
The industry can build agents that can be trusted, verified, and tied to a real identity. This is exactly what makes it possible to keep optimizing for agent compatibility without giving up the relationship underneath it. Striking this balance ensures that technology serves human interests rather than replacing them. By prioritizing identity early, we are able to safeguard the future of customer trust.
The window is now
Compatibility work tends to move within a single roadmap cycle, while building real identity infrastructure takes considerably longer. That gap is the actual risk here. Whichever layer ships first ends up setting the default the rest of the industry builds around, which means the relationship layer needs to be built alongside the compatibility layer now, rather than retrofitted once the market has already settled into a pattern that leaves it out.
Customer identity is core infrastructure for agentic commerce, a foundational improvement rather than an additional feature. Trusted agent protocols, like those we see at Visa and the newly announced AC2, prove how necessary proactive architecture is as we define how commerce operates for the decades to come.
