Mastercard has added 22 companies to a new Agentic Commerce & Services programme. UAE marketplaces and payment providers should prepare for agent identity and consent controls, but most should not rush into a full rollout yet.
Mastercard’s move is a signal to prepare, not a reason to launch AI-led checkout immediately. The company has welcomed 22 companies into its inaugural Agentic Commerce & Services programme, covering agent platforms, AI checkout, identity, payment intelligence, fraud controls, reconciliation and merchant automation. For most UAE marketplaces and payment providers, the sensible action is to map the risks and data requirements first, then test only where the customer journey is easy to control. (mastercard.com)
The programme matters because it is aimed at the infrastructure around agent-led buying, not just chatbots. Mastercard’s list includes Firmly, which helps merchants sell through AI agents; Skyfire, which provides verified agent identity and payment credentials; PayOS, which assesses the card, bank account, business, person, transaction and agent before money moves; and Simetrik, which focuses on transaction-level reconciliation. Other participants address fraud, dispute resolution, catalog readiness, agent connectors and automated business processes. (mastercard.com)
The important shift is from “can an AI agent place an order?” to “can every party prove what the agent was allowed to do?”
Traditional online checkout assumes that a person is browsing, choosing and pressing the purchase button. Agentic commerce changes the sequence. An AI agent may search, compare products, build a basket and initiate payment on a customer’s behalf. Mastercard’s wider agentic-commerce work focuses on registered agents, secure credentials, verified user intent and consent before an action is taken. (mastercard.com)
Its Agent Pay Acceptance Framework describes a model in which agents are registered and verified before they transact. It also refers to agentic tokens, purchase-intent data, consumer identity and an audit trail that can support payment decisions and disputes. Mastercard says merchants may be able to recognise trusted agents at the content-delivery-network layer, with existing checkout forms still used for payment submission. (mastercard.com)
That creates several practical questions for UAE operators:
These are operating and control questions, not merely app-development tasks. A business reviewing its ERP and automation setup should include order status, inventory, refunds, reconciliation and approval records in the discussion. A new AI shopping interface will not fix fragmented back-office data.
Marketplaces should treat agent access as a new channel with its own permissions. Start with read-only product discovery or assisted shopping. Keep high-risk actions, such as changing delivery details, applying unusual discounts or placing large business orders, behind explicit approval. The exact control model will depend on the marketplace, payment provider and agent protocol, so do not assume that one Mastercard announcement creates a single market standard.
Payment providers should watch the identity and intent layer closely. Mastercard is working across several protocols and says the aim is to support clear user intent, secure credentials and verifiable agent identity. That points towards interoperability, but it also means standards are still developing. Providers should monitor the official Mastercard Start Path announcement, then ask vendors how they handle registration, authentication, tokenisation, disputes and data retention. (mastercard.com)
For merchants, product data is part of readiness. An agent cannot recommend or buy reliably if titles, prices, variants, stock, delivery terms or returns information are inconsistent. Businesses planning a new online channel can review this UAE ecommerce setup guide, but the guide will not settle the emerging agent-protocol questions.
Yes, but narrowly. Create an internal inventory of checkout systems, payment flows, bot controls, consent records, product data and reconciliation processes. Ask your payment provider whether it has a documented approach to trusted agents and what evidence it expects from merchants. Run a controlled experiment only if you can limit the products, customers, permissions and payment values involved.
Not yet is the honest answer for a full agentic-commerce rebuild. Mastercard’s programme is an important market signal, but the participating companies cover different layers of the stack and the standards are still being shaped. Do not buy a large platform, rewrite checkout or promise autonomous purchasing to customers solely because 22 startups have entered a Mastercard programme.
The header should use Mastercard’s own digital illustration of a smartphone with holographic shopping icons from its Agent Pay page, with this article’s headline set over the image as a card. It shows the practical link between AI-assisted product discovery and payment, rather than using a generic fintech photograph. (mastercard.com)
Document your current checkout, product data and payment controls, then identify one low-risk journey for a supervised test. If the review exposes gaps in order, stock or reconciliation data, ERP and automation support is the relevant place to start.
Paknology has a commercial interest when that review leads to ERP, automation, ecommerce, website or app work. A cheaper or simpler option may be better if your systems are already joined up and you only need a written control checklist or a small pilot from your existing provider.
Sources
Book a free consultation and get a clear roadmap — from company formation to a fully automated digital operation.