Earlier this year at the National Retail Federation (NRF) Big Show in New York, everything was about AI—from the exhibit hall to the breakout sessions, and conversations at the coffee break areas. One couldn’t get away from it. And it wasn’t just about predictive AI and generative AI. The show was flooded with a newer version of the technology: agentic AI.
Agentic AI refers to autonomous systems that can reason, plan, and take multi-step actions to achieve goals without constant human oversight. Sounds scary? It is. But the implications for retail are boundless.
Aside from work force automation and intelligent support (think of an AI agent working on its own to find solutions to a retailer’s problems), agentic AI can also be used as personal shoppers. For retail, Google’s agents serve as conversational agents that can understand user intent, analyze images, and make personalized recommendations.
But in the rush to deploy these robots into the market, industry experts are warning of potential mishaps as well as privacy and policy issues that may arise.
Balancing AI with the human touch
In a recent report from VoCoVo, titled “In-store Intelligence: AI’s Role in Retail’s Human Touch,” researchers said as stores become increasingly connected and data-centric, “AI is reshaping how customers shop, colleagues work, and retail leaders plan for the future.”
In a survey of 250 retail executives and 500 consumers, the authors of the report noted an acceleration of AI adoption. Amid that ramp up of investments, researchers also revealed some significant tension. The report noted that “while retailers are urgently adopting automation to address labor pressures, rising theft, and operational costs, consumers remain guided by trust, transparency, and comfort.”
The report revealed that shoppers want faster, easier experiences, “but not at the expense of human connection or privacy. This gap between technological ambition and emotional readiness is becoming a central issue for U.S. retailers.”
The survey showed that 43.6 percent of retail executives polled said they plan to implement AI within the next 12 months while 28.1 percent already use it, and 12.7 percent said they are not considering AI at all. The top reasons for deploying AI include reducing checkout time (38.9 percent), improving inventory visibility (27.8 percent), and reducing labor burden (22.2 percent).
The research also showed that nearly 80 percent of shoppers polled said they are unsure how AI is used and expressed discomfort “when it feels like surveillance.”
What is the role of AI in retail?
Martin Smethurst, chief customer officer at VoCoVo, a communications solution provider for retailers, told Sourcing Journal that the research showed a growing gap between retail AI ambitions and consumer trust. “While 94 percent of retailers see AI as a vital tool for operational efficiency—and most shoppers appreciate the convenience offered by automation—our research shows that confusion and anxiety around data use persist.”
Smethurst said the survey results also show that shoppers still rely heavily on human team members for assistance. “Help finding products (38.66 percent), understanding product information (37.63 percent), and resolving issues at checkout (30.27 percent) are key touchpoints where shoppers most depend on [store associates] for help, reinforcing that store staff remain central to the customer journey,” Smethurst said.
With AI shopping assistants, there’s also the issue of trust. While there are guardrails in place, industry experts say AI agents could make mistakes with purchases and other tasks, with the financial burden of those mishaps falling on the consumer.
Smethurst said the success of AI in retail will depend not only on capability, but also on trust. “By distancing themselves from the liability of AI errors, and by viewing AI purely as a strategic necessity, retailers run the risk of eroding the trust that sustains customer loyalty,” Smethurst said.
At VoCoVo, Smethurst said the company believes AI “should accelerate, connect and enhance, making shopping more convenient, faster and more human. AI is most effective when it empowers colleagues to deliver better service—not when it creates a friction-filled barrier.”
Who pays for AI mistakes?
Alix Gallardo, co-founder of useinvent.com, a platform where retailers build AI assistants, told Sourcing Journal that “no retailer wants to be solely responsible for the financial risks of AI errors, especially when their competitors aren’t. That’s why we’re seeing more retailers develop their own native AI assistants, allowing them to set controls, monitor outputs, and intervene when necessary.”
Gallardo said there’s been hiccups with Walmart’s “Sparky” and Amazon’s “Rufus,” and said this is also why OpenAI offers a white-glove solution “tailored for the retail sector, helping companies implement robust AI systems with built-in risk mitigation.”
The use of AI agents also goes beyond the transactional. Gallardo said Lowe’s recently launched “Ask Mylow,” for example, offers not just a list of products with prices, but also detailed step-by-step how-to advice on projects, “helping customers with home improvement solutions to make more informed decisions and gain a comprehensive view of the entire customer journey.”
Gallardo said retailers should be careful about policy changes regarding AI assistants. “Brand trust is built on the implicit promise that a retailer has your back; a terms-of-service clause that says ‘if our AI buys the wrong thing, that’s on you’ signals the opposite,” Gallardo said. “Even if errors are rare, the knowledge that you’re unprotected changes the psychological calculus of using the tool.”
Balancing innovation and accountability
Assaf Feldman, chief technology officer at Riskified ( the fraud protection company used by brands such as Aldo, Prada, and others, said that simply “authenticating the identity behind an AI agent won’t stop chargebacks,” and noted that “the industry must shift to protocols that capture verifiable intent, proving the customer actually authorized the agent’s action.”
Feldman said retailers are navigating a delicate balance between innovation and accountability as AI agents enter commerce, even as consumer expectations remain divided. According to data from Riskified, more than half of consumers believe the AI platform “should be accountable for unauthorized purchases, while 23 percent point to the retailer itself.”
“From here, we expect to see a flood of non-fraud chargebacks from customers claiming that’s not what I wanted my agent to buy, which often morphs into friendly fraud,” Feldman said. “This highlights how intent is tied to the dispute and chargeback process.”
Feldman said retailers risk losing the “direct relationship with their customers, which erodes brand loyalty and lifetime value.” Feldman said moving forward requires safeguards such as verifiable intent, “where a customer’s approval is recorded, encrypted and saved and for merchants to actively invest in their own native AI while ensuring platforms implement risk correctly from day one.”
If all this sounds like a complex, technological policy quagmire, it is.
Jackie Swanson, managing partner at Gartner Consulting, expects terms and conditions policies to be more standardized next year. And it is all about money.
“The economics are straightforward: retailers want to capture the efficiency and conversion gains of agentic commerce without absorbing the financial risk of AI errors at scale,” Swanson said, adding that there will likely be “standard boilerplate [policies] across mass merchants and grocery within the next 12 months. The outlier will be whichever retailer is bold enough to guarantee its AI’s accuracy and uses that guarantee as a competitive differentiator.”
When asked if purchasing policies that hold the customer liable for AI agent mistakes can sour the relationship between a brand and the consumer, Swanson said, “It can, but the real story here is that consumer trust in AI-driven commerce is on a spectrum and it’s going to evolve.”
“Right now, we’re in the earliest innings,” Swanson said. “Most shoppers have never delegated a purchase to an autonomous agent, so the tolerance for error is essentially zero. A single wrong order attributed to an AI assistant can feel like a betrayal, especially when the retailer’s own terms say the mistake is yours to own. That”s a difficult starting point for building confidence.”
Swanson said over time, as these AI tools improve in accuracy and consumers develop a working understanding of what agentic shopping can and can’t do, ”the trust calculus will shift. We’ve seen this pattern before, like when consumers were deeply skeptical of storing credit cards online, of one-click purchasing, of auto-replenishment subscriptions. Each of those required an adoption curve where early friction gave way to comfort through repeated positive experience.”
Regarding other implications of AI policies, there is also regulatory scrutiny to be considered. For example, the EU’s proposed AI Liability Directive “would place the burden on the deployer, not the consumer, in high-risk scenarios,” Swanson said, adding that U.S. regulators, state attorneys general, and the FTC “are watching how agentic commerce unfolds and consumer-unfriendly terms could accelerate intervention.”
Swanson said these new systems also risk creating a divided experience. While tech-literate consumers might navigate the pitfalls successfully, more vulnerable groups—such as the elderly or those less comfortable with technology—are left wide open to errors and exploitation.
However, there is a massive opportunity for a retailer or brand to lead with integrity. The first retailer to stop making excuses and actually guarantee their AI’s accuracy—taking financial responsibility for its mistakes—will win a level of consumer trust that no marketing campaign can buy.
Rachid Wehbi, founder and CEO of Sell The Trend, an AI-powered ecommerce platform that is designed to help entrepreneurs discover winning products, analyze trends, and grow profitable online stores, expects retailers will be slow in AI adoption as shoppers feel they are taking on the risk with AI policies.
“There is also the issue of consent, transparency, risk, liability, and responsibility for biased, incomplete, or commercially driven recommendations,” Wehbi said. “The retail AI sector winners will be those who combine smart AI with customer-first accountability, automation, and uncomplicated reversals.”
