We believe in showing rather than telling. So we built it for ourselves first: an AI-powered chat assistant on blazeforce.nl that greets visitors immediately, asks the right questions and: when the visitor is ready: creates a lead that goes directly into our CRM.
This is not a generic chatbot rented from a SaaS platform and dropped into an iframe. It’s a fully custom WordPress plugin, built on the OpenAI API, with deliberate choices around privacy, language and user experience. And it works.
The context: why we built this for ourselves
Our website attracts visitors with an interest in digitalisation, CRM and custom software. Some of them are well into their exploration: they know what they’re looking for. But another group is still orienting: what exactly does BlazeForce offer? Does this fit our situation? What does something like this cost?
That second group was missing a good first step. A contact form is too much of a threshold. Scheduling a call even more so. What we wanted was a low-friction conversation: available at any time, in the visitor’s language, responding to the page they happened to be on.

What the assistant does
The chat assistant is always available via a widget at the bottom right of the page. It introduces itself, asks what the visitor is looking for and guides the conversation from orientation to intake. What sets it apart from a standard chatbot:
- Multilingual: the assistant automatically detects whether the website is in Dutch or English and adapts. If the visitor writes in English while the site is in Dutch, the assistant switches with them.
- Page context: the assistant knows which page the visitor is on and adapts its opening question and answers accordingly. Someone on the Salesforce services page gets a different introduction than someone on the homepage.
- Intake mode: for specific pages, a “gated” mode is available where the assistant asks for a name and question before providing substantive answers. This filters out low-intent visitors and produces qualitatively better leads.
- Contact form with AI summary: when a visitor leaves their contact details, the assistant automatically generates a summary of the conversation. That summary: including context and any specific questions raised: goes along as a note with the lead in our Nextcloud CRM.
- Feedback module: with each response, the visitor can give a thumbs up or down. That feedback is logged and gives us insight into answer quality over time.
The technical choices
We deliberately chose a plugin that stays fully under our own control. The API key is stored server-side, never in the frontend. Messages are processed through a secured WordPress endpoint: including nonce verification and rate limiting to prevent abuse.
The assistant is built on the OpenAI Responses API using GPT-4o mini as the model. That’s a deliberate choice: fast, cost-efficient and more than capable for this type of conversation. The full conversation history is sent as context with each request, so the assistant can maintain coherence without losing the thread.
Privacy and GDPR compliance were part of the design from the start. The assistant explicitly requests consent before capturing personal data, displays a clear link to the privacy policy and doesn’t store chat histories in the browser or on the server.
From chat to CRM: the lead flow
What makes the assistant genuinely valuable is what happens behind the scenes when a visitor submits their contact details. It’s not just name and email that get captured: at that moment the assistant generates a concise AI summary of the conversation. What was the question? Which product or service was relevant? What level of urgency had the visitor indicated?
That summary is included as a note with the lead automatically created in our Nextcloud CRM. So every follow-up conversation starts with context: not a blank slate.
Result: better leads, lower threshold
Since launching, we’ve seen visitors who previously left without any interaction now starting a conversation. Not every conversation ends in a lead: that’s fine. But the leads that do come in via the assistant are noticeably higher quality than submissions via the standard contact form. They contain more context, come from visitors who have already thought more carefully about their question and are consequently easier to follow up.
What this could mean for you
We built this primarily for ourselves: but the plugin is configurable and deployable on any WordPress website. Whether you’re a service provider wanting to receive better-qualified leads, or an organisation wanting to serve visitors better outside office hours: the approach is scalable and fully adaptable to your brand, tone of voice and processes.
Curious about what this could do for your website? The chat assistant in the bottom right of this page is ready to start the conversation.