Across more than two decades of e-commerce, websites have been built mainly for two audiences: people and search engines. Customers visit a website to find a product, read the information, compare, add to the basket and complete the transaction themselves. Google crawls the content, ranks the results and sends the user to the website.
The rise of the AI agent is changing this familiar structure. In the near future, buyers will not necessarily carry out the whole journey themselves. They can hand the task to an AI assistant: find a suitable product, compare the options, check price and stock, explain the policy, prepare the basket, even help complete the transaction.
At that point a website needs more than a friendly interface or good search optimisation. It must also be clear enough, accurate enough and connected enough for an AI agent to understand the data and carry out the actions it is permitted to take. This is the foundation of the idea of being agent-ready.
The buying journey is starting outside the website
Previously, someone wanting to buy an air purifier might search on Google, open several websites, read reviews and compare specifications themselves. With AI, the user simply describes the need: a room of 30 m², quiet operation a priority, reasonable filter replacement costs and a budget under 10 million dong.
The AI can analyse the request, rule out unsuitable products and propose the closest few options. By the time the user reaches the website, most of the research and the formation of the buying decision has already happened.
Data published by Salesforce on 28 July 2026 shows the number of people starting their shopping journey through an AI conversation rose 200% year on year. Traffic from AI conversations to commerce websites grew between 150% and 428% across the quarters measured. Conversely, the share of consumers discovering products through traditional search engines fell 15% between August 2025 and May 2026.
The research was carried out with 3,450 commerce professionals, 4,689 consumers and behavioural data from more than 1.5 billion shoppers in 37 countries. Only 28% of commerce organisations currently use agentic AI, but another 44% expect to adopt it within six months. That suggests buyer behaviour is changing faster than most businesses are ready for. Salesforce, 28/07/2026
AI is moving from advising to taking part in the transaction
In September 2025, OpenAI introduced the ability to buy directly within ChatGPT for users in the United States. The initial programme supported products from Etsy sellers and was planned to extend to more than a million Shopify merchants. This is one of the clearest signals that AI is shifting from a tool that supplies information to an intermediary layer that can support a transaction. OpenAI, 29/09/2025
On 24 March 2026, OpenAI extended the Agentic Commerce Protocol to help merchants share their catalogue and promotions with ChatGPT. Target, Sephora, Nordstrom, Lowe’s, Best Buy, Home Depot and Wayfair are among the retailers that have integrated for product discovery.
Data from the Shopify Catalog is also connected to ChatGPT, so that products from millions of merchants can appear more fully and accurately within shopping conversations. Rather than pointing the user to a list of links, the AI can present products visually and compare prices, ratings and key features within the conversation itself. OpenAI, 24/03/2026
Commerce through AI remains at an early stage, however. OpenAI has shifted its focus from instant payment towards letting merchants use their own checkout experience. That suggests the market is still finding the balance between convenience, payment security, control of data and the relationship between brand and customer.
The long-term trend is nonetheless reasonably clear: AI will take an increasingly deep part in discovering, comparing, selecting and preparing transactions. A share of the visitors to a website in future may no longer be the buyer directly but an AI agent acting on their behalf.
Agent-ready is not simply adding a chatbot
A website with a chatbot is not necessarily agent-ready. A chatbot is only a conversational interface. Agent-ready is the state in which a business’s data and processes can be understood correctly by an AI system, retrieved at the right moment and used to carry out a permitted action.
A salesperson can understand that “the Pro version in blue” is a specific variant even if the naming in the system is inconsistent. They can also check stock in the internal software themselves, remember a new promotion from a message from their manager or explain a returns policy that happens to be stored as an image.
An AI agent cannot work reliably from data as scattered as that. If the same product has different names on the website, on Shopee and on TikTok Shop, the AI may treat them as three separate items. If the promotional price on the product page differs from the catalogue data, the AI cannot tell which is correct. If the warranty policy appears only inside an image, an external system may not be able to read and interpret it fully.
Agent-ready is therefore first and foremost a data problem. Every product needs a stable identifier, a clear name, complete attributes and accurate relationships between variants. Price, promotions, stock, shipping charges, warranty and returns all need to be updated consistently.
The next step is the ability to connect and to act. Depending on how mature the business is, the AI can check product availability, propose options, create a basket, draft an order, look up order status or pass a complex case to a member of staff.
Why are so many websites not ready?
Most e-commerce websites today are optimised to look good to the user. The data behind them, however, is usually stored across many different systems: the website, the point-of-sale software, marketplaces, Excel, the warehouse and the knowledge of individual staff.
Salesforce records that only 27% of organisations say their customer data is fully unified across sales, service, marketing and commerce. Among businesses selling across multiple channels, only 2% report no significant break in the chain.
The two most common problems are inconsistent pricing and promotions, at 41%, and stock that is not synchronised in real time, at 40%. These have long been familiar operational problems, but the arrival of the AI agent makes their consequences more direct.
A buyer can spend a little more time checking, or call a member of staff. An AI agent has to decide on the data it receives. When it cannot verify price, stock or return conditions, the agent may drop the product from its list of recommendations from the outset.
The business can therefore lose the customer before the buyer has ever visited the website.
From a race for ranking to a race to be chosen by AI
Traditional SEO focuses on being found. Agent-ready widens the problem to being understood, trusted and chosen by AI.
In the old model, businesses tracked keyword rankings, traffic and conversion. In the new model, part of the consideration happens inside ChatGPT, Gemini, Copilot, Perplexity or another AI assistant. The customer may only reach the website after the AI has narrowed hundreds of products down to two or three.
This does not mean SEO will disappear. On the contrary, expert content, a clear website structure and brand credibility remain the sources by which AI verifies information. But businesses will need to go further than keyword optimisation.
Google has long recommended using structured data for products and a Merchant Center feed so that the system understands price, availability, shipping and returns policy accurately. In the context of agentic commerce, these components do more than make a search result look better. They are becoming distribution infrastructure through which machines can recognise and assess a product. Google Search Central
In the years ahead, businesses may need to track new measures as well: in how many purchase-intent queries the brand appears, how accurately the AI interprets the information, what share of AI-referred visitors reach the basket and how much revenue AI has assisted.
Your own website can regain its strategic role
The growth of social media and the marketplaces once led many businesses to treat their website as a secondary channel. But in the era of the AI agent, the business’s own website can return to the centre.
On a website it owns, a business can standardise its data on its own terms, design the advisory process, connect stock and define which actions the AI is permitted to take. The website becomes the brand’s official data source and, at the same time, a point of transaction that does not depend entirely on the distribution algorithm of an outside platform.
This matters particularly in sectors with many specifications or where advice is needed before purchase, such as electronics, furniture, building materials, fashion, cosmetics, medical equipment and B2B products. The more complex the catalogue, the greater the value of clear data and needs-based advice.
An agent-ready website does not replace Shopee, TikTok Shop or Facebook. It plays the role of data and experience hub, while the platforms continue as channels for reaching customers. The advantage lies in maintaining one consistent source of information across the whole system.
Agent-ready does not mean giving AI unlimited access
On 4 August 2026, a United States federal appeals court lifted a temporary order that had barred Perplexity from using an AI agent to help users buy on Amazon. It is regarded as the first US federal appellate ruling to consider an agent accessing an online platform on a user’s behalf. The main case continues, however, and the outcome does not create a freedom to harvest data on any platform. Reuters, 04/08/2026
For Vietnamese businesses, it is important to distinguish clearly between a website you own and a third party’s system. On Shopee, TikTok Shop, Facebook or other platforms, accessing data and taking action must still follow the API, the terms of use and the permissions the platform grants.
A durable agent-ready system must have clear permissions, keep an activity log, require confirmation for important actions and hand over to a person when a situation falls outside its scope. The more an AI can do, the higher the requirement for control.
Where should a business start?
A business does not need to build an AI agent capable of completing every transaction straight away. A more realistic goal is to check whether the current data is enough for an AI to find the right product, quote the right price and stock, explain the policy correctly and then take the customer to the basket.
Preparation can start with the group of products that generates most of the revenue. The business needs to standardise the catalogue and its variants, synchronise price with stock, publish policies clearly, add structured data and identify a few low-risk actions to automate.
In the early stage, AI can support product discovery, compare options, answer questions, recognise buying intent and prepare a draft order. Payment, order changes, complaint handling and decisions with significant consequences should still have a confirmation step from the customer or a member of staff.
The starting point is not choosing the most powerful AI model. The starting point is building a data source reliable enough for customers, staff and AI to use alike.
Agent-ready may become the new standard for websites
Mobile-ready was once a competitive advantage before it became a default requirement. SEO was once a specialist activity before it became an indispensable part of any website. Agent-ready will most likely follow the same path.
In the early phase, an agent-ready business has the advantage of being understood and recommended more accurately by AI. Once AI agents become a common buying channel, that state may shift from an advantage to a basic condition of taking part in the market.
The website of the future therefore needs more than to be attractive, fast and SEO-ready. It also needs consistent data, the ability to connect and readiness to serve a buying journey in which AI takes part from the discovery of a need through to the transaction.
The strategic question is no longer simply whether the business should add a chatbot. The more important question is:
If a customer hands the job of finding and choosing a product to an AI agent, is your website clear and credible enough for that agent to choose it?
T3SKY works alongside businesses to build websites, standardise commerce data and develop a selling experience ready for the era of the AI agent.
