
AI Agents in Online Shopping: Goldman Sachs Sees Visa and Mastercard as Potential Beneficiaries
This article was created with the help of artificial intelligence.
Key Takeaways
- Goldman Sachs estimates that consumer spending of around 2.6 trillion US dollars in the US alone falls into categories highly likely to be suited to agentic commerce.
- Every increase of 2 percentage points in agent-based penetration of card-based offline commerce could lift e-commerce growth by around one percentage point.
- Banks currently wrongly decline payments worth around 30 billion US dollars worldwide each year as fraudulent.
- According to an analysis from 28 April 2026, 43 percent of retailers were testing AI shopping agents at that point.
- Goldman Sachs expects the rollout process to take three to five years or longer.
Artificial intelligence is fundamentally changing online shopping: AI agents are increasingly making purchasing decisions for consumers – from automatically reordering household goods to booking rides. Goldman Sachs Global Investment Research estimates that consumer spending of around 2.6 trillion US dollars in the US alone falls into categories that are highly likely to be suited to so-called "agentic commerce" – defined as low-risk repeat purchases. The analysts see this development as a structural shift in retail that affects payment flows, advertising budgets and fraud prevention alike.
Payment networks as clear beneficiaries
Visa and Mastercard are among the potential winners of this upheaval. More online transactions – including small purchases that were previously made in cash or offline – could generate additional revenue for the payment networks. Goldman Sachs calculates: every increase of 2 percentage points in agent-based penetration of card-based offline commerce could lift e-commerce growth by around one percentage point.
Visa is already developing concrete solutions for automated payments. With "Visa Intelligent Commerce", the company is building payment data, controls, authentication and protection mechanisms directly into automated purchasing processes in order to handle AI-initiated transactions securely and reliably. The announcement was made on 4 September 2026.
Payment processors such as Stripe, Adyen and Checkout are also identified as beneficiaries of growing transaction volumes. The central thesis: value increasingly follows consumers' purchase intent. As this intent is in future expressed less through search engines, online shops or apps and more through AI agents, advertising budgets, payment flows, identity verification and merchant fees will shift accordingly.
Trust determines market success
Mastercard research indicates that the winners will not be the companies with the most grandiose AI promises, but those that make autonomous payments secure, regulated and useful in everyday life. Consumers use agent-based shopping primarily for everyday, recurring tasks – particularly household supplies, where agents replenish stocks within limits set by the consumer at approved merchants. This everyday use creates habits, which is why the first payment experience is decisive for long-term acceptance.
Goldman Sachs names five key capabilities that will determine success or failure: sales strength, consumer trust, merchant connectivity, transaction infrastructure and the ability to capture commercial intent. Large marketplaces that sit at the centre of every transaction are also classed as likely winners.
Fraud prevention as a bottleneck
Traditional rule-based fraud detection systems are reaching their limits in the new environment. Transactions are increasingly shaped by contextual signals, consumer behaviour and AI-driven commerce flows. Banks currently wrongly decline payments worth around 30 billion US dollars worldwide each year as fraudulent – a friction that will intensify in a world where AI agents initiate purchases without traditional card-present signals.
Google has launched an initiative called the "Agent Payments Protocol", supported by more than 60 organisations. In parallel, fraud prevention companies are developing tools to identify and stop malicious agents.
From household needs to luxury goods: a staggered rollout
Goldman Sachs expects low-risk, less complex categories to be taken over by agents first. Products with relatively low brand loyalty – such as paper goods, cleaning products, packaged food and groceries in general – can be reordered by agents based on historical purchase data, prices, inventory levels and delivery speed. Ride services, restaurant reservations and delivery services also meet the requirements for early adoption.
The hardest purchases to automate are those with high personalisation and high error costs: shoes, jewellery, certain clothing, luxury goods, travel, hotels and flights require more sophisticated authorisation, confirmation and liability rules. Overall adoption depends on transaction complexity and the cost of wrong decisions; the process is likely to take three to five years or longer.
Retail pilot projects are increasing
According to an analysis from 28 April 2026, 43 percent of retailers were testing AI shopping agents at that point. The affected sectors range from e-commerce and digital advertising to payments, identity verification, fraud prevention and cybersecurity.
Structural change with winners and losers
The migration of purchase intent from search fields, retail websites and apps toward AI agent interfaces shifts not only payment flows but also advertising budgets and merchant distribution fees. Companies that control transaction infrastructure – above all Visa and Mastercard – are positioning themselves for a central role in this new commerce architecture. Whether they can sustain that role over the long term depends on their ability to integrate security, usability and trust into autonomous payment processes.
Sources
- Goldman Sachs Maps the Agentic AI Shopping Era: $2.6 Trillion in Consumer Spending Set to Be Rewritten, Payment Networks Emerge as Winners
- Consumer AI agents could upend online shopping - but who really wins?
- Consumer AI agents could upend online shopping - EUROPE SAYS
- When AI Agents Start Doing the Shopping
- 43% of Retailers Are Piloting AI Shopping Agents | PYMNTS.com
- What will make consumers trust AI shopping agents? Startup founders explain | Mastercard US
- Enabling AI agents to buy securely and seamlessly | Visa
- AI Shopping Makes Context the New Payments Currency | PYMNTS.com