Agentic commerce is buying done by AI agents: software that researches suppliers, compares offers and, increasingly, completes the purchase for a person or a company. If an agent cannot read your prices, stock, delivery terms and policies, it cannot shortlist you, so your job is to publish those facts in a form machines can read and trust. Below: what it means for B2B and B2C sellers, which platforms and protocols are live as of October 2026, what agents look for, the risks, and a 30-day checklist for Egypt and the Gulf.
What is agentic commerce, in plain terms?
Agentic commerce is any purchase where an AI agent does part of the buying work a person used to do. The person sets the goal ("find three food-grade packaging suppliers that deliver to Jeddah within a week") and approves the result. The agent does the reading, the comparing and, where payment rails allow it, the checkout.
A search engine gives your buyer ten links. An agent reads those pages itself and returns a shortlist of two or three. If your page is vague, you do not rank lower. You are left out.
Our pillar post on AI trends for 2027 names AI agents as buyers as one of seven shifts. This is the practical version.
Why does agentic commerce matter for B2B sellers right now?
Because your buyers already use AI to build shortlists, and Gartner expects agents to sit in the middle of almost all B2B buying within two years. Gartner predicts that by 2028, 90% of B2B buying will be AI agent intermediated, pushing over $15 trillion of B2B spend through AI agent exchanges.
The habit is already forming. In a Gartner survey of 645 B2B buyers published in May 2026, 45% said they used GenAI during a recent purchase, mainly to gather information on vendors and products. Your website is one of their sources, and a model increasingly reads it before a person does.
The Gulf is moving early. When Visa launched its Agentic Ready program in the UAE in May 2026, it cited research with Fast Company Middle East in which nearly 60% of UAE businesses expressed strong interest in, or were actively exploring, agentic commerce use cases.
Which agentic commerce platforms and protocols are live as of October 2026?
The payment rails are live in pilots and early rollouts; checkout inside an AI chat is still being worked out. Based only on each company's own announcements:
| Launch | Who | Launched | Status, October 2026 |
|---|---|---|---|
| Instant Checkout and Agentic Commerce Protocol (ACP) | OpenAI with Stripe | Sept 2025, US | Since March 2026, merchants can use their own checkout; ACP now carries product feeds into ChatGPT |
| Agent Payments Protocol (AP2) | Google with 60+ organizations | Sept 2025 | Open protocol; coming to Google products, starting with Gemini Spark |
| Universal Commerce Protocol (UCP) | Google with Shopify, Etsy, Wayfair, Target, Walmart | Jan 2026, US | Powers checkout in AI Mode and Gemini; Canada and Australia next, then the UK |
| Visa Intelligent Commerce | Visa | Apr 2025 | Live transactions since late 2025; first regional use with Aldar in the UAE |
| Mastercard Agent Pay | Mastercard | Apr 2025 | First transaction outside the US in the UAE; Agent Connect for merchants announced Sept 2026 |
The most useful signal is OpenAI's change of course. It launched Instant Checkout in ChatGPT in September 2025 with US Etsy sellers and the open Agentic Commerce Protocol, co-developed with Stripe. Six months later it said the first version "did not offer the level of flexibility" it wanted. Merchants can now use their own checkout and share product feeds through ACP, while OpenAI focuses on product discovery.
Google is building both layers: the Agent Payments Protocol, launched with more than 60 organizations including Mastercard and PayPal, and the Universal Commerce Protocol, which covers discovery through post-purchase support. Its May 2026 shopping update extends both.
The card networks are making agent payments traceable, and the Gulf hosted early live cases. After launching Visa Intelligent Commerce in April 2025, Visa reported hundreds of agent-initiated transactions by December. Mastercard's first Agent Pay transaction outside the US took place in the UAE with Majid Al Futtaim, and Visa's first live regional implementation let an Aldar customer pay property service charges through an AI agent.
The lesson: the fight happens at discovery, long before payment. And no protocol fixes vague product data. Protocols carry your data. They do not write it.
What does an AI agent need from your business before it can recommend you?
It needs facts it can read without guessing: price with currency, stock or lead time, specifications, delivery terms, return and warranty policies, and proof of who you are. Anything locked in a PDF, an image or a "contact us for pricing" form is invisible to it.
Structured data states those facts in a format machines agree on. Google's merchant listing documentation requires a Product with a name, an image and an Offer carrying price and priceCurrency, and recommends availability, shipping details, return policy and GTIN. Add Organization markup so an agent can confirm you are a real company.
Access matters as much as markup. OpenAI says its OAI-SearchBot crawler surfaces websites in ChatGPT search and recommends allowing it in robots.txt. Make sure no firewall or CDN rule blocks it by default.
For B2B, add what procurement agents compare: minimum order quantity, lead time, payment terms, certifications and delivery countries. Service firms should publish packages with "from" prices.
How do you check whether your website is ready for agentic commerce?
Run your top 20 product or service pages through this table. Every row an agent cannot answer is a reason to pick someone else.
| What agents look for | Where it lives on your site | How to fix it |
|---|---|---|
| Price and currency | Product and service pages | Show prices or "from" ranges in EGP, SAR or AED; mark up with Offer |
| Stock or lead time | Product pages, product feed | Sync from your inventory or ERP, not by hand |
| Specifications | Product pages | Move specs from PDFs and images into HTML tables |
| Delivery terms | Shipping page, product pages | Zones, times and costs per country; add shipping markup |
| Returns and warranty | Policy pages | One plain page per policy, linked from every product |
| Company identity | About page, footer | Organization markup: legal name, address, phone |
| Product feed | Google Merchant Center, ChatGPT | One clean feed generated from your catalog |
| Live data access | API or MCP server | Read-only endpoints for price, stock and quotes |
| Speed and crawl access | Whole site | Fast server-rendered pages; allow search crawlers |
The last two rows matter most in B2B. An agent that queries your price and stock through an API gets a current answer, not a stale page. That is the job of the Model Context Protocol, explained in What Is MCP? We run our own CMS at HBS through an MCP connection, so AI works inside the real system. Your catalog deserves the same.
How does agentic commerce connect to GEO and AI search visibility?
They are two halves of one job: GEO gets your brand named in the AI answer, and agent readiness lets the agent act on it. A model that cites you but cannot read your price still sends the buyer elsewhere to finish.
The competition for that slot is large. Google says its Shopping Graph holds over 60 billion product listings. The winners are complete and consistent: the same facts on your site, feed, marketplaces and directories, in Arabic and English.
Our GEO guide for 2026 covers the content side. Our Advanced SEO solutions cover the technical side: structured data, feeds, crawl access and speed.
What are the risks of selling through AI agents, and how do you keep the relationship?
The real risks are price wars, fake agents and losing direct contact with the buyer. Each has a practical answer.
- Price wars. When two offers look identical on paper, an agent compares price. Make your other strengths machine-readable: delivery time, warranty length, local support and Arabic service are facts an agent can weigh.
- Fake agents. Bad bots can pose as shopping agents. Visa built its Trusted Agent Protocol to help merchants tell trusted agents with commerce intent from malicious automation. Ask your payment provider which agent standards it supports, and keep human approval on large or unusual orders.
- Losing the relationship. If the agent owns the conversation, the buyer may never see your site. Keep checkout on your own store where the platform allows it, as ChatGPT now does, and earn loyalty after the sale.
People still close deals. In the same Gartner survey, 69% of B2B buyers prefer to validate AI-generated insights with sales reps. Put a named contact and a WhatsApp number on every key page. The agent builds the shortlist. Your people close it.
What should Egyptian and Gulf businesses do in the next 30 days?
Fix the facts first, then the structure, then the access. The order we recommend:
- Days 1 to 3: test yourself. Ask ChatGPT, Gemini and Perplexity, in Arabic and English, to shortlist suppliers in your category and city. Note whether you appear and what they get wrong.
- Days 4 to 7: pick 20 pages. Your highest-revenue products or services.
- Week 2: publish the facts. Prices or ranges with currency, stock or lead time, delivery terms by country, returns and warranty, identical in Arabic and English. Move specs out of PDFs into HTML.
- Week 3: add structured data. Product, Offer and Organization markup, checked with Google's Rich Results Test.
- Week 3: open the doors. Check robots.txt, firewall and CDN rules so Googlebot and OAI-SearchBot can reach your pages. Fix the slowest ones.
- Week 4: build one feed. Generate a Google Merchant Center feed from your catalog or ERP so price and stock update on their own.
- Week 4: plan live access. If you sell B2B, scope a read-only API or MCP endpoint for price, stock and quotes.
- Day 30: test again. Re-run the day 1 prompts and compare.
UAE sellers should also ask their bank whether it is part of Visa's or Mastercard's agent programs.
Is your business ready to be chosen by an AI agent?
Answer that for your top 20 pages, not your homepage. Protocol names will keep changing. The data under them will not: a clear price, a real stock number, honest delivery terms and a company an agent can verify. I would rather see a business publish clean facts for 20 products this month than wait a year for the perfect agent strategy.
Ahmad, Founder of HBS Group
Want to see how AI agents read your business today? Book a 15-minute call and we will run the shortlist test with you and show you the gaps. Or explore our Advanced SEO solutions for structured data, product feeds and AI search visibility.
