With OpenAI, Mastercard and Visa setting the stage for a future where AI agents could take control of online shopping on behalf of consumers, the e-commerce industry is bracing itself for a significant shift in its relationship with customers. Retail Systems news editor Alexandra Leonards explores the opportunities and risks of agentic AI for the retail market as the technology accelerates.

Major developments in agentic commerce in recent months have demonstrated that a reality where AI agents can shop, compare and buy goods on behalf of consumers is closer than many anticipated.

In October, Mastercard completed a transaction using an agentic token that enables AI agents to execute payments on behalf of users. Shortly afterwards, Visa unveiled a foundational framework for agentic commerce designed to establish secure communication between merchants and AI agents during transactions.

Arguably the most significant move so far has been made by PayPal, which recently followed in the footsteps of Shopify, Etsy and Walmart in partnering with OpenAI. Last month, the company announced plans to embed its digital wallet into ChatGPT to connect tens of millions of merchants to the platform.

The partnership is timely, with new research from B&Q and Screwfix owner Kingfisher revealing that more than a quarter of UK adults (28 per cent) are already comfortable allowing AI shopping assistants to make purchases autonomously without their explicit permission.

These developments offer a glimpse into the near-term future of e-commerce, characterised by streamlined transactions and reduced friction in the checkout process. Whilst agentic commerce promises retailers new sales channels and higher conversion rates, the technology's potential to fundamentally transform the merchant-consumer relationship presents considerable risks if not approached with strategic planning and robust safeguards.

Redefining the merchant-consumer relationship

Traditional e-commerce follows a linear journey: customers search, visit multiple sites and compare products before making a purchase decision.

"Merchants 'own' the customer through their websites, loyalty programmes and data collection," Dr Eri Bertsou, assistant professor of political science at the University of St Gallen, tells Retail Systems.

In an LLM-powered shopping environment, this journey collapses into a single step, with the AI service capturing the data, preferences and loyalty that merchants have traditionally cultivated through direct customer relationships.

Bertsou, who specialises in behaviour, democracy and decision-making, points out that whilst some data will always remain with the merchant, the quantity of data points will "decrease sharply" with the advent of agentic commerce.

Elias Ghanem, global head of the Capgemini Research Institute for Financial Services, says that shopper behaviour once anchored in brand loyalty will shift towards algorithm-driven decisions.

He anticipates that differentiation based on price, reviews, delivery speed and inventory availability will become more influential for consumers than brand loyalty.

Dr Andrea Barbon, assistant professor at the Centre for Financial Services Innovation at the University of St Gallen, warns that if orders originate through AI chat interfaces, merchants will lose critical customer intelligence from the front end.

"For instance, what did the user search for to get to the product? Which other products did they consider before reaching the final decision? How much time elapsed from the first search to the actual purchase?" he continues. "This valuable information is at risk of slipping away from merchants in favour of chat agents."

Xavier Sheikrojan, director of risk intelligence at commerce protection platform Signifyd, highlights potential efficiency gains for retailers.

"When AI agents handle the mechanics of a transaction, the checkout process becomes almost invisible, which means that payments are faster and abandonment rates could fall dramatically," he explains. "For merchants, that means higher conversion rates and potentially increased repeat business as customers grow used to a seamless shopping experience that feels effortless.”

If retailers can successfully identify and track orders made by AI agents, they may unlock significant long-term data opportunities. Once merchants understand transaction patterns initiated by agents, they can develop more sophisticated strategies for fraud detection, dynamic pricing and customer experience optimisation.

Competing for attention

As the merchant-consumer relationship evolves, a parallel relationship will develop between retailers and AI agents.

James Fry, head of enterprise product at payments processor Worldpay, says that merchants who once relied on direct site visits will now compete to be selected by agents, rather than simply attracting human attention.

"Ownership of the relationship becomes shared across platforms, agent providers and merchants unless retailers adapt their data, product feeds and identity strategies," he continues.

Signifyd's Xavier Sheikrojan says that retailers are already beginning to adapt to this emerging dynamic with AI.

"We're beginning to see sellers refine their product listings so they’re to be understood and prioritised by AI systems, not just human shoppers," he continues. "As agents become the main interface between consumers and brands, retailers will be competing for algorithmic selection rather than clicks or impressions."

Sheikrojan adds that being 'AI-readable' is becoming as critical as visual appeal.

"Retailers will need to ensure their product data is sufficiently structured and transparent enough for agents to interpret confidently, particularly as peak season approaches," he says. "This Black Friday and Christmas won’t be a full picture of what's to come, but it will offer the first indication of how AI might start shaping discovery and decision-making."

He emphasises that retailers must strike a balance: continuing to optimise for human shoppers whilst establishing the infrastructure for AI agent interactions with their platforms.

Capgemini's Elias Ghanem argues that as merchants optimise their offerings for algorithmic evaluation rather than human psychological appeal, this "levels the playing field" by diminishing the significance of marketing budgets and brand recognition. Instead, operational and technical capabilities will emerge as key differentiators.

"The implication for merchants is clear: operational excellence is the difference between success and failure," he tells Retail Systems. "This is why merchants are prioritising payment success rates, service reliability and integrated fraud management as top criteria when selecting technology providers."

Dr Eri Bertsou suggests that merchants may compete for AI agent selection in ways analogous to search engine optimisation strategies for Google rankings.

However, she notes this will be a gradual transition because consumers currently prefer maintaining control over product selection criteria, payment information and final purchase decisions.

Worldpay's James Fry emphasises that retailers must produce accurate, structured and machine-readable product information at scale whilst maintaining synchronisation between inventory systems and delivery commitments.

“Legacy systems often fragment pricing, promotions and stock data across channels, so achieving the real-time fidelity agents need will require platform and data modernisation,” he continues. “Merchants will also need to decide how to prioritise agent relations, whether to pay for visibility, and how to measure returns when attribution becomes opaque.”

Managing fraud risk

Whilst the benefits of agentic commerce are evident — faster and more seamless transactions — the migration of payment choice and authentication upstream to the agent's environment poses challenges. Merchants could lose visibility into payment verification processes and prioritisation of payment methods, potentially impacting processing costs, conversion rates and fraud exposure.

When transactions occur through AI agents, retailers may lack the data necessary to effectively assess risk or understand purchase intent.

Sheikrojan explains that without access to browsing data, device signals and, in some cases, clear records of consumer consent, detecting fraud and resolving disputes becomes significantly more challenging.

Agentic AI can also circumvent controls designed to identify human behavioural patterns because it can replicate the digital footprints that shoppers typically leave behind.

Agents can solve CAPTCHAs, spoof device identities and fingerprinting, complete checkouts remotely, or utilise stored tokenised credentials—capabilities that allow them to bypass traditional fraud prevention measures such as high-value transaction thresholds.

Sheikrojan notes that fraudsters are already exploring synthetic agents that mimic legitimate human behaviour, whilst legacy fraud detection systems struggle to distinguish between beneficial and malicious automation.

“As a result, merchants may see a rise in successful fraudulent orders, more chargebacks and disputes, weaker evidence in chargeback investigations, and greater margin erosion from returns and wrongly applied promotions,” he continues. “Crucially, liability and resolution get more complex when the merchant never saw the shopper’s journey or the authentication steps used.”

Addressing these challenges will require retailers and payment platforms to develop new verification and trust frameworks that preserve accountability alongside innovation. This becomes increasingly urgent as agentic commerce adoption accelerates.

Entering the mainstream

PayPal's partnership with OpenAI represents a watershed moment in the evolution of agentic commerce.

Luca Russignan, deputy head of the Capgemini Research Institute for Financial Services, explains that amidst parallel developments from Mastercard and Visa, the PayPal-OpenAI collaboration distinguishes itself through potential scale and ecosystem openness.

Rather than operating as a closed ecosystem limited to a single retail environment, this model is platform-agnostic, potentially connecting tens of millions of merchants across the internet.

Embedding PayPal's digital wallet directly into ChatGPT effectively compresses the entire shopping journey—from discovery through payment—into a single conversational interaction.

“It marks a potential shift from ChatGPT as a conversational platform to a commerce marketplace and cements PayPal as the underlying platform to enable payments,” continues Russignan.

Sheikrojan observes that whilst this development benefits consumers by simplifying online purchases to conversational requests, it fundamentally alters the competitive landscape for merchants.

"We can expect this to be the first of many such partnerships, demonstrating how rapidly payment experiences are converging with product discovery," he says.

Russignan argues that from a trust and consumer protection perspective, this represents a strategically sound and well-timed initiative.

“Consumers are already comfortable buying through platforms like Amazon because there’s a clear trust layer for payment protection and chargebacks,” he explains.

Integrating PayPal's established safeguards into ChatGPT extends this confidence framework to AI-driven transactions.

“The challenge ahead is adapting chargeback and dispute processes for a world where AI, not the consumer, triggers the payment, redefining what ‘authorisation’ really means,” he continues.

Whilst adoption will likely begin gradually, constrained by merchant integration requirements and consumer confidence levels, acceleration could occur rapidly once conversational commerce demonstrates both frictionless operation and robust security.

“As we’ve seen with the surge in digital wallets and mobile payments, the tipping point will come when convenience and trust converge,” he adds.

Agentic commerce will likely gain substantial traction once consumers recognise they can seamlessly transition from browsing options to completing purchases within a single conversation.

“The PayPal-OpenAI partnership doesn’t start that journey, but signals that the race is now well underway,” he says. “If successful, this could redefine the checkout experience, from a distinct step at the end of the journey to an invisible moment within it.”

Building trust for the future

Dr Andrea Barbon anticipates that over the medium to long term, the industry will see expanding merchant ecosystems accepting agentic commerce-enabled purchases.

“Further, ‘agentic commerce protocols’ will be released by AI companies, including schemas for consent, KYC handoffs, dispute data, and refund hooks," he continues.

Russignan suggests that the central design challenge will be orchestrating seamless user experiences.

“The winners will be those who make the shift from search to purchase feel instantaneous and intuitive, while maintaining clarity, consent, and brand control throughout,” he says.

Critical to this will be embedding trust, transparency and verification mechanisms that ensure all stakeholders maintain confidence in agent-driven transactions.

Achieving this requires focus on what Dr Eri Bertsou describes as the technical foundations of trust.

“Algorithmic transparency in plain language for users to understand will be key,” she says. “The core challenge is that for this system to work, everyone needs to feel confident about what's happening with their money.”

Specifically, consumers need to understand the reasoning behind AI agent recommendations — whether a particular product or merchant was genuinely optimal, or whether commercial relationships influenced the selection.

“You need to know if they are getting commissions for placing products,” she says. “Additional controls, spending limits and authentication processes for merchants all need to be built into the model.”

Bertsou emphasises that broader questions surrounding accountability, privacy and data ownership require external regulation by governmental institutions and regulatory agencies, rather than self-regulation by AI developers.

“Considering the number of intermediaries that need to be trusted, this might not be an easy task,” she says.

Consumers would need confidence in the AI agent itself, the merchants it selects, the underlying algorithms, the payment systems, the security infrastructure and the overarching regulatory framework.

“Nevertheless, adoption of AI agent purchases might outpace stated trust, especially if the AI agents provide clear, immediate value like convenience, savings, and better decisions,” she says.

Whilst widespread trust may not materialise as rapidly as technology optimists anticipate, Bertsou notes it's also not as distant as sceptics suggest.

“The change will be gradual over the next decade, starting with strong brand ties and only limited adoption for low-value purchases, and increase once social proof kicks in, positive feedback loops and more tech comfortable youngsters replace the more reluctant older consumers,” continues Bertsou.

Worldpay research indicates that UK shoppers expect approximately seven per cent of their total online purchases to be conducted via AI agents by 2030, potentially representing up to £29 billion in online spending.

However, most Britons remain reluctant to cede complete purchasing control to AI, with 60 per cent indicating they want to review every transaction before authorisation.

In an era where data privacy dominates consumer concerns and AI scepticism persists, retailers bear responsibility for ensuring that innovation and convenience don't compromise accountability, transparency and security.

As AI agents continue disrupting established merchant-consumer relationships, retailers must carefully balance service convenience with meaningful human connection that cultivates customer loyalty, trust and engagement — all whilst developing new differentiation strategies suited to an algorithmic marketplace.

---