Imagine a visitor to Tobago asking an AI assistant to find a guesthouse near the beach for three nights next week, under a set budget, with breakfast included. The assistant checks what is available, picks two options, and, once the visitor says yes, books one. The visitor never opens the guesthouse’s website.
That is not a distant idea. The large AI and payment companies are already building it. For thirty years we designed websites for people. Now we also have to design them for the assistants that act on people’s behalf. The industry has a name for this: agent experience, or AX.
What is AX?
AX (agent experience) is how easy it is for an AI assistant to understand your business and complete a task for a customer: check a price, find a free slot, make a booking, place an order. The term was introduced in January 2025 by Mathias Biilmann, chief executive of the web platform Netlify, who described it as the experience AI agents have as the user of a product.
He placed it in a line that starts with UX. Don Norman coined “user experience” in 1993 to describe everything about a person’s experience with a system. AX asks the same question for a new kind of user: an AI assistant working for your customer.
Is this actually happening yet?
Yes, although it is early and most of it started in the United States:
- Buying inside ChatGPT. In September 2025, OpenAI and Stripe launched Instant Checkout, letting ChatGPT users in the US buy from certain online shops without leaving the chat, and published an open standard (the Agentic Commerce Protocol) so other businesses and assistants can do the same.
- Payments made by agents. The same month, Google announced the Agent Payments Protocol, developed with more than 60 payment and technology companies including Mastercard, American Express and PayPal, to let AI agents pay on a customer’s behalf with proof of what the customer approved.
- Booking through search. In August 2025, Google began testing AI Mode features that search reservation platforms for live restaurant availability and take the customer to the booking page, with local service appointments and event tickets to follow.
- Agents are already visiting websites. In June 2026, BrightEdge reported seeing 88 requests from AI agents acting for users for every 100 visits from ordinary search, with OpenAI accounting for 96% of that agent activity. BrightEdge sells tools for tracking this, so treat the figures as a signal rather than a precise count.
How does a business get ready for AX?
We see it as three stages, each building on the one before.
1. The brain: know your own facts
An assistant can only act on facts that exist somewhere it can read. Prices, availability, what is included, cancellation terms, delivery areas, opening hours. In most businesses these live in someone’s head, a WhatsApp thread or a spreadsheet. The first job is to write them down in one checked place, a Business Knowledge Base, so there is one true version to publish from.
2. The layers: be found and understood
Next, the business has to be found, trusted and understood: listed by search engines, chosen as the answer, cited by AI assistants and described correctly everywhere. These are the layers we explain in GEO vs SEO (and AEO, AIO, SXO). An assistant will not book a business it cannot find or does not trust.
3. AX: be easy to act on
Finally, the assistant has to be able to complete the task. That means the facts are stated plainly, the actions are simple, and there is a route to book or buy that a machine can follow, with a person approving the final step.
Why are UX and AX mostly the same work?
Because what confuses a person usually stops an assistant too. Many AI browsing tools read a page through the same hidden labels that screen readers use for people with visual impairments. OpenAI’s own guidance for its Atlas browser says making a website more accessible helps its agent understand the page and use buttons, menus and forms correctly.
So the practical steps overlap almost completely:
- Facts written as text, not hidden in images, PDFs or “call for rates”.
- Clearly labelled buttons and forms, so both a person and an assistant know what each one does.
- Short, predictable steps to book or buy, without dead ends that only a human could get past.
- Structured data (a small layer of code that labels facts such as prices, availability and opening hours) so machines can read them reliably.
- Accurate listings on the booking and payment platforms that assistants already connect to, such as reservation systems and online shop checkouts.
Good UX was always about removing friction. AX removes it for one more kind of visitor.
Are you ready? An eight-question check
Answer yes or no. Every “no” is a place where an assistant, or a customer, is likely to give up.
- Are your prices, or clear “from” prices, written as text on your website?
- Can a customer see what is available, or how to check, without phoning?
- Are your cancellation, delivery and refund terms written down and easy to find?
- Are your name, address, phone number and opening hours the same everywhere online?
- Can a booking, order or enquiry be completed online in a few simple steps?
- Are the buttons and form fields on your website clearly labelled?
- Are you listed, with correct details, on the booking or ordering platforms your customers use?
- Does a person on your team confirm or review anything that commits money or time?
Many businesses will answer “no” to several of these. That is not a crisis. The same fixes help today’s customers, today’s search engines and tomorrow’s assistants.
Should a small business act now or wait?
Act on the foundations now, and don’t build anything elaborate yet. The standards are new, several are competing, and most of the early features are limited to the United States. But the groundwork (a checked record of your facts, clear pages, simple booking and accurate listings) pays off immediately with human customers and positions you for assistants as they arrive.
This is how we work at UX Caribbean: see where you stand, capture what you know, make it visible and keep it current. Our AI Visibility Audit checks each stage and produces an AI Readiness Report.
Frequently asked questions
What is AX (agent experience)?
AX is how easy it is for an AI assistant to understand your business and complete a task for a customer, such as checking a price, making a booking or placing an order. It is the equivalent of user experience (UX), but for AI assistants acting on a person’s behalf.
What is the difference between UX and AX?
UX is the experience a person has using your website or service. AX is the experience an AI assistant has when it uses them for a person. The two overlap heavily: clear information, labelled buttons and simple steps help both.
Can AI assistants book or buy without visiting my website?
Increasingly, yes. Some assistants can already complete purchases inside the chat through shop and payment platforms, and search tools are testing booking features. Where they do use your website, they read it much like a screen reader does, so clear, accessible pages matter.
What is agentic commerce?
Agentic commerce means AI assistants finding, comparing and buying products or services on a customer’s behalf. Open standards such as OpenAI and Stripe’s Agentic Commerce Protocol and Google’s Agent Payments Protocol are being built so this can happen safely, with the customer approving what is bought.
Do I need a new website for AX?
Usually not. Most of the work is making facts clear, labelling buttons and forms properly, simplifying how people book or buy, and keeping listings accurate. These fixes help human customers first, and assistants follow.
What is the first step?
Write down the facts a customer needs before they buy (prices, availability, terms, hours) in one checked place, and make sure your website states them plainly. Everything else, from search visibility to AI bookings, builds on that record.
Sources
- Mathias Biilmann, Introducing AX: Why Agent Experience Matters, 28 January 2025.
- Stripe, Stripe powers Instant Checkout in ChatGPT and releases Agentic Commerce Protocol codeveloped with OpenAI, 29 September 2025.
- Google Cloud, Powering AI commerce with the new Agent Payments Protocol (AP2), 16 September 2025.
- Google, AI Mode in Search gets new agentic features and expands globally, 21 August 2025. Initially for US subscribers testing it through Google Labs.
- BrightEdge, BrightEdge Finds ChatGPT Represents 96% of Live AI User-Agent Activity, 16 June 2026. BrightEdge sells AI search and agent tracking software.
- OpenAI, Publishers and Developers FAQ (ChatGPT Atlas), accessed September 2026.