Choosing a door means working through dimensions, design and individual options. Jan is configured to support that process through product information and a connection to the webshop. This story looks at the approach and the first recorded activity, with the early stage of the implementation kept in view.
THE IMPLEMENTATION IN NUMBERS
- active conversations
- 19
- AI replies
- 53
- ready knowledge sources
- 1
29 October 2026 · first measured day. Customer channels; no dashboard or prompt tests. Knowledge sources: inventory at snapshot. How we count ↗
01
Choosing a door involves more than one question
Deurwebshop presents doors that visitors can compare and configure to measure. Someone looking for a new door may start with colour or appearance. As the choice becomes more specific, other questions follow: which version suits the opening, what layout is needed and which options belong to the selected model? Each answer can lead to another decision.
Advice needs to keep up with that process. A link to a product page may be a helpful start, but a visitor can also need support halfway through a configuration. At that point, the useful answer connects to choices already made. Asking the visitor to start their search again would lose the context that makes the question meaningful.
Jan has a defined role within this experience: a Dutch-speaking door advisor configured to help visitors choose, compare and configure. The setup connects explanations with the webshop’s available functions. It is a recent implementation. The capabilities described here were checked in the configuration and should not be read as the results of a long-term performance study.
02
Understand the question before searching for the product
A focused question gives the conversation a useful starting point. Is the visitor choosing a door for a particular use? Have they already selected a model? Do they want to compare options or continue configuring? The intended workflow does not send every question straight to the configurator. Jan can first help clarify what the visitor is trying to achieve.
When the relevant functions are available, the connected shop can provide product information. Concrete products and available options need to stay connected. The configuration directs the assistant to use dimensions and choices from the available data. An option that sounds reasonable is not necessarily one that can be selected for a particular door.
At the snapshot, Jan had one processed knowledge source linked to him. That is a starting inventory, rather than a measure of the breadth or accuracy of the advice. For current product choices, the shop connection can also matter. The available knowledge and the functions accessible in the current session together determine what the assistant can support with evidence.
03
A next step that connects to the shopping experience
Consider a visitor who knows the type of door they want but needs help selecting a version. The intended workflow is to clarify that choice, retrieve a relevant product and check the options available for it. The conversation can then connect to a configuration. This illustrates the setup; it is not a report of a particular customer conversation.
The connection also matters when a question concerns a visitor’s own cart or project. Jan is configured to check which actions the website offers for that session before acting. Without the connection, he should not invent the current cart contents or claim a change has succeeded. Explaining an action and completing it are different responsibilities.
Explicit visitor consent is part of the configuration for cart changes. The customer keeps control over the choice being made. Completing a final order or making a payment independently is outside this role. The assistant can therefore support a practical stage of the buying process while maintaining clear boundaries around actions that have consequences.
04
The first measured day: 19 conversations and 53 AI replies
Jan was created on 28 October 2026. On 29 October, up to the snapshot at 18:20 Netherlands time, 19 active conversations and 53 AI replies were recorded on the embedded website channel. Dashboard conversations and prompt tests are excluded. These are early activity figures, rather than a monthly average or evidence of completed purchases.
That distinction matters for a new implementation. A first day can show that conversations are being started and the assistant is providing replies. It tells us less about returning use, seasonal patterns or the proportion of questions fully resolved. Understanding those outcomes requires a longer reporting window and a separate assessment of the conversations.
This story therefore uses the counts that can be verified. A revenue uplift, a savings claim or a satisfaction score would go beyond the available evidence. The first conversations still give the team a concrete starting point for following the implementation and identifying where a follow-up question or a product explanation could be more useful.
05
Developing the assistant around real choices
With made-to-measure products, the quality of a conversation often depends on small details. Does the visitor understand an option? Does the explanation fit the selected model? Is the next step clear? These are useful questions for reviewing the first conversations. They connect improvement to the actual shopping process, rather than simply to the number of answers sent.
Jan’s setup also shows why each business benefits from defining its assistant’s role. Robby at MartijnKozijn is configured mainly to explain and guide. Jan is configured around door choices and functions in the shop itself. Differences in knowledge, tone and available actions follow from the task the visitor is trying to complete.
For other webshops, the practical starting point is specific: choose the process where visitors need help, connect the relevant product data and keep the visitor involved in each change. Deurwebshop is at the beginning of that development with Jan. The first measured day makes the start visible; later snapshots can show how the use develops over time.
About the figures in this story
Source: Corthex analytics, verified snapshot on 29 October 2026 at 18:20 (Europe/Amsterdam). Counts are limited to this assistant and customer channels; dashboard and prompt-test channels are excluded. The final day is partial.
An active conversation contains a visitor or AI message within the period. AI replies are recorded assistant messages, excluding employee messages. Knowledge sources are processed, linked sources at the snapshot. Conversations are not unique customers; channel filtering does not exclude every possible test on a website. These counts do not measure verified resolution, revenue or satisfaction.
Jan was created on 28 October 2026. All measured activity is on 29 October; this is an early implementation story, not a long-term results study.
The described workflow was checked against the assistant configuration. Examples are illustrative; no customer transcripts or quotations are published.
Business information and imagery: Deurwebshop ↗