Bol launches Shopper Agent: what every business can learn

What is Bol Shopper Agent?

Shopper Agent is the personal AI shopping assistant that Bol, the largest online retailer in the Netherlands and Belgium, launched in beta for all customers on 15 September 2026. Instead of typing keywords you describe what you need in plain language, and the assistant asks follow-up questions before recommending products, complete with pricing and delivery times.

You find it in the app and on the website, directly next to the search bar. In Dutch the same feature is called Shophulp.

An example Bol uses itself: you say you are going camping, where you are going, when, and with how many people. You do not get a list of tents back. You get a shopping list built around your situation.

For transparency: our founder Stefan Hoogers advised Bol on building this shopping agent. What follows is shaped partly by what we watched work up close.

Why Shopper Agent is not a normal chatbot

The difference is not the answer. It is the question that comes before it. A classic chatbot waits until you supply enough information and then searches. Shopper Agent asks: how many people, which season, how much experience do you have.

That sounds like a small thing and it is not. Asking questions collects information a search bar never receives. A query like 'tent 4 person waterproof' reveals nothing about this being your first camping trip, that you are travelling with small children, or that it is October.

That is where the gain sits. Not in searching the same catalogue more cleverly, but in gathering context that was always there and that nobody thought to ask for.

The shift: from searching and comparing to personal guidance

Ronald van Rijn, Chief Product, Technology and Data Officer at Bol, describes it as online shopping moving away from searching and comparing towards personal guidance.

It is a sharp observation, because it touches something every business recognises. Customers do not arrive with a product name. They arrive with a situation: a leak, a house move, a present for someone who owns everything already.

Most websites answer that situation with a search box and a contact form. The gap between what the customer knows and what your website asks for is left for the customer to bridge. Often they do not bother, and they leave.

What this means for smaller businesses

The reflex on reading news like this is: good for Bol, but we do not have an engineering team that size. True, and yet the conclusion is not that you should wait.

The principle behind Shopper Agent is not expensive. It is the difference between an agent that guesses and an agent that asks. That second question is the whole thing, and it now ships in off-the-shelf platforms.

In fact smaller businesses hold an advantage a marketplace with millions of items does not: you know your own range and your own customers. A heating engineer knows the three questions that decide whether a boiler fits. A camping shop knows which detail settles the choice.

That knowledge is exactly what an agent needs in order to ask the right follow-up question. Bol had to infer it from data. You already carry it in your head.

What you can borrow without Bol's budget

Start with the questions your own staff already ask. Listen in on the phone for a week or read back your chats. Which two or three questions does a good salesperson always ask before recommending anything? That is your blueprint.

Have your agent ask those questions instead of answering immediately. An agent that asks 'is this a new build or an existing property?' before quoting a price gives a better answer than one that leads with the cheapest option.

Know when to stop asking. This is the part that most often goes wrong. Too many questions and people abandon the conversation, too few and you have rebuilt search with extra steps. Two or three targeted questions is the ceiling for most businesses.

Be clear about what you do with data. Bol uses purchase history and wishlists only when the customer explicitly asks for it. That restraint is not a limitation, it is a reason people are willing to use it at all.

How we solved this in Heyloha

In the Heyloha AI chatbot the agent decides for itself when to ask and when to answer. For a question with a limited set of options it renders choice buttons, so the visitor answers with a click instead of typing a sentence.

When it needs several pieces of information at once, it renders a compact form inside the chat rather than working through them one question at a time. That cuts drop-off and produces cleaner data, because required fields prevent typos in email addresses.

The knowledge comes from your own website and documents, not from a general model. The agent does not guess at answers and says so honestly when it is unsure, rather than inventing something plausible.

And when a situation gets too specific, it hands the conversation to a colleague along with everything already discussed. The customer never has to tell their story twice.

The real lesson of Shopper Agent

The interesting thing about Shopper Agent is not that Bol uses AI. Everyone uses AI. The interesting thing is where they put it: next to the search bar, at the moment someone does not yet know what they are looking for.

That is the moment most businesses lose their customers. Not at checkout, but long before it, when someone arrives with a vague question on a page that only answers precise ones.

Want to see what happens when your visitors meet something that asks the second question? Try Heyloha free for 14 days on your own website. No credit card, live within 30 minutes.

Frequently asked questions about Bol Shopper Agent