Agentic commerce is one of those things: One day it’s on the periphery of your awareness and then, overnight, it’s everywhere all at once.
Of course, that doesn’t just happen. It could have something to do with the fact that in relatively rapid succession OpenAI, Mastercard, Visa, Walmart, Google, Amazon and PayPal with Perplexity announced agentic AI initiatives or enhancements.
And while there has been a lot of talk, it’s been more than just talk. There’s substance: The announcements by some of the world’s biggest companies in commerce and payments have provided a peek into what’s to come — and likely sooner rather than later.
What is agentic commerce and when will it happen?
What everyone is talking about is a subset of the Gen AI revolution: automated AI agents that for now find products and monitor prices for consumers and alert them to make purchases when the prescribed conditions are met.
The announcements have rolled out with equal parts what-is-possible and what-is-aspirational, but a critical mass of experts is sure something is happening — and that the something is significant.
“It’s a pretty big shift in how shopping happens,” says Xavi Sheikrojan, Signifyd director, risk intelligence. “We have these AI-powered agents. They will now soon shop. They will decide. They will even check out on behalf of customers.”
And they will change plenty about ecommerce. Marketers and merchandisers will need to optimize for bots. Fraud and risk teams will need to instantly differentiate between good bots and bad bots — and train themselves, and machine learning models, to interpret new signals characteristic of benign AI agents shopping on behalf of loyal customers.
Kumar
Most likely, says Signifyd Chief Technology Officer Varun Kumar, merchants will need to create parallel online shopping universes — one designed for the way humans like to shop and one designed for the way agentic bots like to shop. Risk teams will need to adopt the same dual vision, knowing that the signals produced by humans and those by bots vary widely. Patterns that signal a malicious bot, for instance, and those indicating a malicious human actor are not the same. The same goes for good bots and good customers.
What are the pros and cons of agentic commerce?
On the one hand, full-on automated shopping bots are the ultimate in frictionless commerce. It’s one step short of “if you think it, you can buy it.” But as with every advancement or technology that makes online shopping easier, agentic commerce also opens up merchants to new fraud vulnerabilities and opens up consumers to new fraud worries.
Grocery is a logical leader in agentic commerce
Consumers are understandably hesitant to blindly hand over their digital wallets to shopping bots. So, we’re likely a ways away from a mass movement to a world where AI agents — guided by prompts on selection, price, quantity and other guardrails — are automatically completing purchases.
In laying out the need for building trust in the emerging agentic channel, Bain & Company produced a survey of U.S. consumers with ROI Rocket. They found that most consumers are familiar with AI, but that only 10% had ever used it to purchase a product.
Most of those, Bain wrote, relied on AI agents to buy groceries or household items. Makes sense, right? Groceries are generally low-risk, low-cost items. Many items on our grocery lists are weekly regulars — bread, eggs (OK, not everything is low cost), milk, toilet paper. Substitutions are usually tolerable — should a roma tomato show up instead of a vine-ripe, no problem.
Apparel shows promise in the AI agent era
Some apparel purchases seem primed for early agentic commerce activity. Certainly not the rare purchase of that perfect outfit for that special occasion, although that day is coming. But most wardrobes have a few go-to garments — polo shirts, socks, yoga pants, workout tops — some versions of which are purchased repeatedly, as needed.
Electronics are likely to cross the agentic chasm in classic fashion
The computer and electronics category sits in an intriguing spot at the dawn of the agentic commerce era. Many purchases involve high-value devices by consumers with fairly specific lists of features and specifications. Those contemplating a completely autonomous shopping journey might hesitate to turn so much over to a machine learning model. But many technology enthusiasts are early adopters and could be ready to go for it, even before the agentic commerce platforms are ready for them. Like so much about agentic commerce in 2025, we’ll have to see how adoption plays out in electronics.
The travel vertical will be another interesting study in agentic commerce
Travel, purchasing airline tickets for instance, is likely to follow a pattern similar to electronics. We’ll likely see frequent flyers and business travelers turning to AI agents early on and less-frequent, leisure travelers watching to see how things shake out.
Beauty and cosmetics, home goods, sporting goods and other ecommerce verticals will be mixed
Consumers shopping in other key retail categories — home goods, beauty and cosmetics, sporting goods and the like — will most likely adapt to agentic commerce on a highly individual level. The categories include products that consumers can be fairly particular about. Home goods are often big ticket items that will be around for a long time, for instance. Some sports equipment requires just the right fit or feel. But a regulation basketball is a regulation basketball. Beauty and cosmetic purchases literally affect the way the world sees us. The safe bet here is that agentic commerce will grow steadily in all categories — the question is how quickly.
How should retailers prepare for the new era
However quickly consumers take to the new form of shopping and however long it takes for the necessary infrastructure to be in place to support full-on automated buying, retailers will be preparing for significant changes.
Xavi Sheikrojan, Signifyd director, risk intelligence, digs into what the evolution of agentic commerce will mean for ecommerce fraud and risk professionals.
For fraud and risk professionals that means adopting solutions and systems that provide commerce protection from bad bots and bad people without turning away orders from legitimate AI agents and legitimate human customers.
“For retailers and for fraud teams as well, it introduces a lot of fundamental shifts in risk, in visibility and in control,” Sheikrojan says. “Traditional fraud prevention relies a lot on end-to-end visibility. So you get a couple of key challenges.”
Maybe more than a couple.
Top eight challenges of agentic commerce for merchants
1. Lack of full visibility
Agentic commerce brings with it a lack of vision into where along the buying journey the key vulnerabilities arise. The introduction of third-party agentic bots potentially makes it harder to know if fraud is occurring at the agent stage or farther downstream on the merchant’s site. Not knowing where the specific vulnerability lies makes it difficult to assign responsibility and accountability.
2. Increased consumer confusion
Relying on an army of shopping agents could mean more chargebacks. For instance, a shopper might see a charge attributed to a purchase made via an AI agent. The statement could list the agent’s platform rather than the merchant involved. Or maybe the bot purchased from a merchant the consumer isn’t familiar with. That could lead to a dispute. Miscommunication could also complicate matters. What a consumer meant to ask a bot to shop for and what a bot actually determines a consumer wants could differ, leading to disappointment and chargebacks.
3. BTO or bot-takeover attacks
Think of BTO as account takeover on steroids. Seizing control of an agentic commerce bot could become the height of online fraud. Controlling the agent would allow fraudsters to race through the web making purchases in the victim’s name.
4. More social engineering and imitation schemes
Phishing and impersonation schemes could increase. Fraud rings might launch social engineering scams in which they send shoppers notices confirming agentic purchases they never made and suggesting they click on a link for more information. Once the consumer clicks, the fraudster makes off with valuable personally identifiable information and credentials.
5. Neutralizing site-level fraud protection
Depending on how and where a transaction actually takes place, bot purchases could circumvent a merchant’s extra site-level fraud controls. Maybe a merchant has created rules to weed out high value transactions or high-velocity activity. But if the purchase doesn’t make a traditional run through that site, those rules are meaningless for that particular purchase.
6. Higher return rates
If the agentic bot is providing product photos and descriptions or sizing guides or reviews, there is a chance some of that information could be inaccurate, depending on the original source. Erroneous information could lead to disappointment and a return and a refund request.
7. Need for new thinking
Fraud and risk professionals live in a world where bots are bad. Bots launch card-testing attacks.They enable mass purchases that fuel unauthorized reselling. Models are trained to deflect bot traffic. Old thinking runs the risk of turning good orders and revenue away — losing the immediate sale and perhaps future sales from that disappointed customer or well-trained bot.
8. Speed and scale in general
In fact, the enhanced automation that shopping agents provide means any scheme a fraudster or fraud ring can come up with can be executed with stunning speed and at breathtaking scale.
What steps are available today for merchants looking toward the future?
The technology and approaches to successfully navigate the agentic commerce era already exist. Not to say that there isn’t more work to be done on solutions that were designed with human shoppers in mind. Consider, however, how solutions like those provided by Signifyd apply.
How Signifyd can help in an agentic commerce world
As AI plays a bigger and bigger role in consumers’ online shopping excursions, AI-driven commerce protection solutions become all the more important. Signifyd’s Commerce Protection Platform combines the scale of data and the sophistication of machine learning models necessary to prepare for the future of commerce. The technology is backed by fraud, risk, data and technology experts who can provide the kind of consultative guidance that every successful retailer will need as the world in which we all operate continues to evolve and innovate.
Understanding identity and intent
Signifyd’s Commerce Network protects merchants based on millions of daily transactions executed on many retail verticals around the world. Driven by AI, it establishes a baseline of behavior that is updated constantly by additional behavioral and transactional signals. To operate successfully in the agentic commerce era, you want visibility into a new baseline of bot behavior that doesn’t resemble human behavior, but doesn’t result in fraud either. That builds an understanding of the identity and intent behind automated purchases.
Consortium data for better insights
Signifyd provides insights into shopping agents’ meta data and it analyzes their AI-driven intent. This gives merchants important context behind a purchase by combining global pattern recognition with individual transaction insights. It analyzes whether a transaction is consistent with past behavior, preferences and authorization patterns or whether the transaction is out of the ordinary. Signifyd’s solutions also differentiate between human-approved automation and malicious bots, using techniques like AI fingerprinting and transaction metadata analysis. By leveraging real-time updates from a global merchant network, intent models continuously learn from new attack patterns and evolving buyer behaviors.
Intelligent returns protection and management
Signifyd can also protect against an increase in returns — whether abusive, fraudulent or excessive — stemming from bot behavior. Decision Center enables merchants to customize the response to return requests depending on the level of risk they entail. Similarly, Instant Refunds provides a way to calibrate the speed with which the refund is approved. A merchant can refund a loyal customer’s purchase immediately or delay a refund for a consumer with a checkered history until the returned product and its condition are verified.
What does all this look like right now — today?
AI-powered commerce protection is more important than ever under the current conditions. Think about how Google has described its AI Mode shopping experience. A consumer selects a product and requests a price. Google alerts the consumer when there is a match. Then the shopper taps “buy for me” on the alert.
“Behind the scenes, we’ll add the item to your cart on the merchant’s site and securely complete the checkout on your behalf with Google Pay,” Google’s blog post announcing the product explained.
Since the purchase is ultimately completed on the merchant’s site, it stands to reason that the merchant’s existing fraud prevention strategy and systems still go to work. But that’s only half the equation.
Humans are vital to agentic commerce success
The bigger concern, given that a bot is playing a role in the buying journey, is that an order will be falsely declined by static and legacy systems that don’t have the intelligence to recognize that the beneficial bot has made other purchases elsewhere on the web without incident. Identifying orders placed by helpful bots requires a broad consortium model that establishes recognizable patterns sorting good bots from bad bots.
What is clear from the current agentic commerce picture is that success will not be a product of machines alone. While bot vs. bot provides a nice storyline, the fact is navigating this new world is going to take a large dose of human intelligence to guide merchants’ strategies and solutions in this brave — and promising — new world.
It’s not too soon for merchants to start assembling those necessary elements now. All signs point to agentic commerce being more on the immediate roadmap than part of a fantastical science fiction future.
Photo by Getty Images
Looking to chart your agentic commerce future? Let’s talk.
