Agentic commerce: The quiet revolution in how we buy

Eman does not remember the last time she manually booked a flight. Her personal AI assistant, linked to her credit cards and loyalty programmes, quietly scans fares, checks her calendar, considers her seat and airline preferences, and then books the optimal option while she sleeps. By the time she wakes up, the boarding pass is in her wallet, the hotel is reserved, and the best card has been chosen for rewards and instalment options. Eman only set the guardrails; the agent did the rest.

This is agentic commerce: the next stage of e‑commerce where AI agents don’t just recommend products; they autonomously browse, compare, decide and pay on our behalf, typically using card rails in the background. Instead of you logging into a site to renew your insurance or hunt for flight deals, your agent continuously monitors prices, inventory, loyalty points and your personal constraints, then executes the transaction through pre‑authorised credentials.

Payment networks and consultancies are already sketching the outlines of this world. They describe a rethinking of shopping itself, collapsing today’s funnel of search, comparison, reviews and checkout into an intent‑driven, machine‑orchestrated flow. Under the hood, networks are building the plumbing so that merchants and issuers can verify that a bot acting on a consumer’s behalf is legitimate and that consent exists for each transaction.

The scale of the prize explains why agentic commerce has vaulted to the top of board agendas. McKinsey estimates that agentic commerce could generate up to 1 trillion dollars in orchestrated US retail revenue by 2030, and between 3 and 5 trillion dollars globally. BCG’s work on payments suggests that AI agents could shape more than half of online purchases in advanced markets over the next decade, as they increasingly initiate or complete transactions rather than merely recommending options. For card schemes and banks, that implies a future where transaction volumes shift from sporadic human checkouts to always‑on machine‑to‑machine flows riding their rails.

For consumers like Eman, the appeal is obvious. Delegating routine purchases – groceries, household supplies, subscriptions, utility payments – to an agent could reclaim hours each week while ensuring continuous optimisation of price, quality and timing. Mature agents will not only choose merchants but also dynamically select payment methods across cards and accounts to maximise rewards, minimise fees and smooth cash flow.

Networks and banks also stand to gain. When card credentials become deeply embedded in agent ecosystems, those who control standards, identity and dispute rules can anchor trust and capture new revenue from fraud analytics, risk scoring and data‑driven offers. Issuers that make their cards “agent‑friendly” – with good APIs, flexible controls and rich data – can defend top‑of‑wallet status and gain more granular insight into customer behaviour. For merchants, reliable agentic flows can mean stickier relationships, more predictable demand and lower acquisition costs.

“Agentic commerce will only work if consumers feel that their digital twin is both smarter and more ethical than their human self,”  said Dimitrios Dosis, president of Mastercard for Eastern Europe, Middle East & Africa, to me on a recent podcast. Dosis captured the central paradox of AI‑driven shopping: for us to hand over our wallets to algorithms, we must first hand over our trust. If people fear their agent is quietly optimising for someone else’s margin, or locking them into a closed ecosystem, adoption will stall.

One concern is algorithmic lock‑in. As agents learn from our interactions and negotiate with other agents, switching costs will rise sharply. Once your preferences are deeply embedded in a particular ecosystem – a mega‑platform, wallet or network – rival merchants may find it hard to get visibility, much as many today struggle with search rankings and social feeds.

Fraud and liability are another minefield. If an agent overspends, misinterprets a preference or is manipulated by a malicious bot, who is responsible – the consumer, the bank, the platform or the merchant? In an agentic world, identity, consent, reversibility and dispute handling must become first‑class design principles. Consumers will also need simple controls: caps on agent spending, category‑level permissions, and one‑click ways to revoke consent.

There is also a social risk. If affluent consumers and sophisticated businesses gain access to more powerful agents first, they may enjoy better prices, credit terms and product access than those still trapped in manual channels. The same technology that democratises access to information could, paradoxically, widen the gap in financial outcomes between the digitally fluent and the rest.

For the GCC – and the UAE in particular – agentic commerce represents both a natural next step and a strategic choice. The region combines high smartphone and card penetration with ambitious national AI strategies and regulators comfortable with sandboxing new models.

On the infrastructure side, the pieces are falling into place. Mastercard has launched Agent Pay in Dubai with Majid Al Futtaim, enabling consumers to use AI assistants to browse, compare and purchase items such as VOX Cinemas tickets, in a pilot witnessed by the UAE’s Minister of State for AI. This builds on the country’s advanced digital ID, high card penetration, and open attitude to cross‑industry partnerships.

For Gulf banks and retailers, the message is clear: agent‑readiness will require structured product data, robust APIs, tokenised credentials and experiment‑friendly regulation – all areas where the UAE has a head start but cannot be complacent.

For now, agentic commerce still feels like a glimpse of tomorrow. But for Eman – and soon for many of us – it may quietly become the default way we shop, travel and pay. The real question is not whether AI agents will buy for us, but whose values they will embed, whose interests they will serve, and which societies will shape the rules before the code hardens.

- The writer is the host of the popular “Money Majlis” podcast. He was previously the global head of retail banking and wealth management at a leading regional bank.