Your next customer might not be human. Is your business ready to sell to an AI agent?

Picture a normal Tuesday in 2027. A customer in the UAE tells her phone: "Book a deep clean for the flat this week, under AED 400, someone with good reviews." She never opens your website. An AI agent reads five providers, compares prices and slots, and pays one of them.

If your prices live in a PDF, your availability lives in someone's WhatsApp, and your policies live nowhere, that agent can't choose you. Not because you're worse. Because it can't read you.

This isn't science fiction. The first pieces are already live in the UAE.

What is agentic commerce, and is it live in the UAE?

Agentic commerce is when an AI agent searches, compares and pays on a person's behalf, within limits that person sets. It is live in early pilots in the UAE: Mastercard ran Agent Pay with Majid Al Futtaim in November 2025, and Visa ran an agentic payment with Aldar in December 2025. Visa opened its Agentic Ready programme in May 2026.

• November 2025: Mastercard ran its first Agent Pay transaction outside the US in the UAE, with Majid Al Futtaim. An AI agent bought a VOX Cinemas ticket on a customer's behalf (Majid Al Futtaim).
• December 2025: Visa and Aldar completed what they called the region's first end-to-end voice-enabled agentic payment: a customer paid real-estate service charges through Aldar's AI agent (Visa).
• May 2026: Visa launched its Agentic Ready programme in the UAE, with banks and fintechs including Emirates NBD, Mashreq, ADCB, Wio, Tabby and Ziina testing agent-initiated payments. Research by Fast Company Middle East with Visa found nearly 60% of UAE businesses surveyed were strongly interested in or actively exploring agentic commerce (Visa).

The payment rails are being built now. The question for every business is what sits on the other side of them.

Will AI agents replace my customers' buying decisions?

No, not soon. Gartner predicts over 40% of agentic AI projects will be cancelled by the end of 2027, and it warns about old chatbots rebranded as agents. But the payment rails are being built now. The businesses that gain will be the ones that agents can read easily, not the ones with the flashiest agent.

Gartner's prediction (June 2025) puts the cancellations down to cost, unclear value or weak risk controls, and it calls the rebranding problem "agent washing" (Gartner). The same release predicts that by 2028, a third of enterprise software will include agentic features and at least 15% of day-to-day work decisions will be made autonomously.

Both things are true. The hype is ahead of reality, and the direction is clear.

How does an AI agent see my business?

An agent reads what is structured and public. It uses on-page prices, a booking calendar with an API and a written refund policy, not a PDF price list, availability that lives in WhatsApp, or rules nobody wrote down. If a search engine struggles to parse your offer, so will an agent, and it will pick a competitor.

• Price: Hard to read: in a PDF or an image; easy to read: structured on the page or in a feed.
• Availability: Hard to read: "call us" or a WhatsApp chat; easy to read: a booking calendar or inventory with an API.
• Policies: Hard to read: unwritten, or buried; easy to read: a clear page: cancellation, refunds, what's included.
• Language: Hard to read: machine-translated Arabic; easy to read: Arabic written natively.
• Payment: Hard to read: bank transfer details by message; easy to read: a checkout that supports tokenised payments.

What should I fix before agents arrive?

Start with five fixes. Put your prices and packages in structured form. Make availability visible to systems. Write down your policies. Clean up your data before you add AI. Write Arabic natively, not through machine translation. None of these needs an AI agent of your own, and each one also helps the human customers you already have.

1. Make your offer machine-readable

Prices, packages, service areas, delivery times, opening hours: in structured form on your site (schema markup, a clean product feed, or an API), not only in images and PDFs. If a search engine can't parse it, an agent can't either.

2. Put your availability somewhere a system can see it

Agents book what they can confirm. A booking calendar or inventory system with an API beats a phone number every time.

3. Write down your rules

Cancellation, refunds, warranty, what's included. Agents compare policies as well as prices, and they skip what they can't verify.

4. Get your data in order first

Roland Berger found that 4 in 5 GCC organisations have an AI strategy, but only 34% have the enterprise-wide data foundation to scale it (Roland Berger). For SMEs the gap is wider: a 2026 du and Huawei study of 648 UAE SMEs found only 15% use AI and analytics platforms today (Zawya). Agents don't fix messy data. They expose it. Mapping where your data is incomplete or locked away is usually the right first step.

5. Be readable in Arabic, not just translated

A large share of agent requests in this region will be in Arabic, including Gulf and Egyptian dialects. Product names, policies and descriptions written natively in Arabic will be understood and matched better than machine-translated ones. It's how we've worked on Arabic-first products since 2021.

Does UAE e-invoicing push in the same direction?

Yes. UAE e-invoicing becomes mandatory from January 2027 for businesses with revenue of AED 50 million or more and from July 2027 for those below it, and a scanned PDF does not count as an e-invoice. Egypt already runs e-invoice and e-receipt systems. Tax authorities and AI agents both need structured, connected business data.

The phased dates are set by the UAE Federal Tax Authority (FTA), and the Ministry of Finance is clear that PDFs, images and scanned copies are not e-invoices; the data has to be structured (UAE Ministry of Finance). Egypt has been on that road for years. Whether the reader is a tax authority or an AI agent, the message is the same: your business data needs to be clean, structured and connected. (E-invoices themselves go through an accredited service provider; the work on your side is getting systems and data ready.)

Where should a business start this quarter?

Ask one question: if an AI agent tried to buy from us today, where would it get stuck? Walk through finding, comparing, checking availability, paying and getting a receipt. Every step where a human has to read a PDF or answer a message is a fix to rank by effort and value. Do the cheapest high-value one first.

You don't need to build an agent. Each place it gets stuck is a fix that also helps your human customers.

How we'd approach this

We'd start small: pick the three processes closest to revenue (often enquiries, quoting and booking), walk each one through as if an agent were the customer, and note every point where it needs a human to read a PDF, answer a WhatsApp message or check a spreadsheet. Then we'd rank the fixes by effort and value, and do the cheapest high-value one first. Most of the early fixes are plain data and integration work, not AI.

Want a second pair of eyes on your processes? See how an AI & Automation Assessment works: up to three processes, a ranked list of where AI and automation pay back, and a recommended first step. Prefer to talk it through? Book a free 30-minute intro call.

Your turn: when do you think AI agents will make a meaningful share of purchases from businesses like yours: already, by the end of 2027, in 2028–2030, or not in your industry? Email us at hello@fritill.com.

Fritill is a technology venture studio and digital transformation partner working with founders and organisations in the UAE, Saudi Arabia and Egypt.

Sources checked 3 November 2026: Majid Al Futtaim press release (Nov 2025); Visa Middle East press releases (18 Dec 2025, 20 May 2026); Gartner press release (25 Jun 2025); Roland Berger Middle East (Feb 2026); du/Huawei SME study via Zawya (2026); UAE Federal Tax Authority and Ministry of Finance e-invoicing pages. Predictions are labelled as predictions.