
Two kinds of AI now sit between shoppers and products, and they are easy to mix up.
The first works on your store. A shopper asks your assistant which serum suits oily skin and gets an answer based on your catalogue.
The second works somewhere else. A shopper tells an AI agent: "Find me a waterproof commuter backpack under €120 that arrives this week, and order it." The agent decides which store gets the order, and the shopper may never see your product page.
Both get called "AI shopping". For a merchant, they are different problems with different to-do lists.
Conversational AI shopping helps the shopper decide on your store. Agentic AI shopping lets an outside agent decide and act for the shopper.
For the wider background on agentic commerce and the protocols behind it, read our explainer on agentic commerce. This article focuses on the difference between the two models and what to do about each.
Conversational AI shopping: your assistant, your store
Conversational AI shopping means shoppers talk to an AI assistant on your website in their own words instead of relying only on search, filters and long product pages.
They ask the questions they would ask a sales associate:
- •"Which of these is a good gift for my dad?"
- •"What is the difference between these two models?"
- •"Is this suitable for sensitive skin?"
- •"Which size should I choose?"
A good assistant asks a clarifying question when the request is vague, recommends products from your catalogue, compares options, explains details and answers questions about shipping or returns based on your published policies. Where the integration supports it, it can also add a product to the cart.
The shopper stays in charge. They read, compare, decide and complete checkout themselves. The assistant works like a knowledgeable salesperson on the shop floor: it guides, but it does not buy.
This is also what separates an AI shopping assistant from a scripted FAQ bot. We cover that in the difference between AI chatbots and AI assistants.
Agentic AI shopping: an outside agent buying for the shopper
Agentic AI shopping starts with delegation. The shopper gives an AI agent a goal and permission to act within limits.
The agent may then:
- •interpret the goal and constraints
- •search products across several merchants
- •compare price, delivery times and policies
- •build a cart
- •ask for confirmation, or act within limits the shopper approved in advance
- •start or complete checkout
The defining shift is delegated execution. The agent does not just advise; it carries out steps of the purchase on the shopper's behalf.
Just as important is where it happens. The agent runs in an AI assistant, search interface or platform that you do not operate. It reads your store through product data, feeds and interfaces rather than through your page design.
Platforms and payment providers are building the infrastructure for this, for example Google's Universal Commerce Protocol and the Agentic Commerce Protocol developed by Stripe and OpenAI. These standards are still evolving, and what is available differs by platform and market.
The difference at a glance
| Conversational AI shopping | Agentic AI shopping | |
|---|---|---|
| Who runs the AI | The merchant | A third party: an AI assistant, search engine or platform |
| Where it happens | On your storefront | Outside your store |
| Shopper's role | Asks, compares and decides | Sets a goal and approves, or pre-authorises |
| Scope | Your catalogue | Potentially many merchants |
| Checkout | Completed by the shopper | May be started or completed by the agent |
| What the AI reads | Your catalogue and configured knowledge | Structured product data, feeds and protocols |
| Merchant control | High: you set sources, tone and actions | Limited: the agent applies its own criteria |
| What you learn | What shoppers ask before they buy | Often only the order itself |
Why the difference matters for merchants
Conversational shopping works on the traffic you already have
Many shoppers do not leave because checkout is broken. They leave because they are unsure: they cannot find the sizing, compare two similar products or tell whether something fits their need.
An on-site assistant works in that hesitant middle part of the journey, while purchase intent is still there. The conversation stays on your store, so you keep control over the product context, brand experience and customer relationship.
You also get something new: a record of what shoppers ask. Repeated questions show which products confuse people and which information is missing from your product pages.
Skincare brand Elinor uses ShopAssist this way, as an online beauty advisor. In the first three months, orders assisted by ShopAssist made up more than 12% of its webshop revenue. Read the Elinor customer story for details.
Agentic shopping may change where discovery happens
If more shoppers ask an external agent to find and buy products, your website is no longer the only place where the decision happens.
That raises different questions. An agent compares you with every other merchant selling something similar, using structured criteria such as price, delivery and return terms. Brand storytelling and on-site browsing have less influence on that comparison. It is also less clear who owns the customer relationship and who is responsible if the agent buys the wrong item.
Agentic shopping should not be treated as automatic new demand. It may become a new channel, and it may also replace traffic from channels you already understand.
What to do for conversational shopping
- •Put guidance where hesitation happens. Product pages, category pages and search are where shoppers stall. That is where an assistant earns its place.
- •Connect it to your real sources. Answers can only be as good as the catalogue, attributes and policies behind them.
- •Set action boundaries. Recommending, comparing, explaining and adding to cart are a sensible scope. Checkout stays with the shopper.
- •Read the conversations. Treat repeated questions as research, not just support work, and fix the product pages they point to.
What to do for agentic shopping
- •Make product data machine-readable. Clear titles, variants, identifiers, attributes, prices, currencies and availability. Agents cannot reliably compare products from vague marketing copy.
- •Give policies one source of truth. Shipping, returns, warranty and refund rules should live in one clear place, not across several pages.
- •Ask your platform before building anything. Most stores do not need a custom buying agent. Check which agent-related features your eCommerce platform and payment provider support.
- •Decide what an agent may do. Define when you require confirmation and how errors, returns and disputes are handled.
- •Watch without chasing every announcement. The protocols are moving quickly. Understand the direction, but put your effort into the foundations.
The same foundations serve both
The two models have different owners and live in different places, but they depend on the same groundwork.
If a product page confuses a human shopper, it is probably weak input for an AI agent too. If a return policy is spread across three pages, neither a shopper nor an agent will apply it correctly. If attributes are missing, every comparison becomes unreliable.
A conversational assistant helps you find these gaps faster, because shoppers point them out in their own questions. Fixing them improves the store for people today and makes it easier for agents to read later.
Where ShopAssist fits
ShopAssist is on the conversational side. It is an AI shopping assistant that runs on your own storefront and works with Shopify, WooCommerce, BigCommerce and Magento.
It is designed to help shoppers find, compare and understand products based on your catalogue and configured knowledge. It answers questions about store policies, adds products to the cart where supported and shows you what shoppers ask before they buy.
ShopAssist is not an autonomous buying agent. It does not buy across merchants, run checkout inside external AI agents or implement an agentic commerce protocol. Its job is to make your store easier to understand, compare and buy from, which is also the groundwork for agent-led shopping. You can see how plans are structured on our pricing page.
The takeaway
Conversational AI shopping improves decisions on your store. Agentic AI shopping moves some decisions away from it.
For most brands, the right first step is not to build an agent. It is to make your catalogue, policies and buying journey clear enough for a shopper to decide without help, and to put a good assistant there for when they cannot.
If you would like to see how that works on your own product pages, book a demo.
Ready to turn visitors into buyers?
See how ShopAssist can boost your conversion rate and revenue.

