BlazeForce AI Chat is the AI intake assistant running on our own website, and the same plugin we deploy for clients. We already wrote about what the assistant delivers in leads and conversion. This article shows the other side: not a script with an API key bolted on, but a fully self-built WordPress plugin. Here is how it’s structured, what it can do, and how security is handled, in a way that shows how BlazeForce approaches this kind of custom work.
How it’s built
The assistant is built as a standalone WordPress plugin, independent of any theme or specific website, so the same foundation can run on multiple client sites without starting from scratch each time. The code is split into clear parts: a settings screen for administrators, the actual chat logic, and the connection to the AI provider.
In practice it works like this: the visitor chats in the browser, the message travels through the website’s own server to the AI provider (OpenAI), and the response comes back along the same route. The AI key itself never reaches the browser. The assistant also automatically pulls context from the page the visitor is on, so a question on a services page gets a different kind of answer than a question on an article page.
For lead follow-up, the plugin is connected to both email and our own Nextcloud environment, with its own fallback mechanism: if one route fails, the other doesn’t automatically fail with it. The full source code is under version control (Git), and updates are rolled out through a repeatable deployment process with a technical check right before a change goes live. A change to the live site is therefore never a manual action without a safety net.
What it can do
- Bilingual intake chat (Dutch/English), with a tone of voice, company narrative and pricing approach that are all fully configurable.
- Page-aware answers: the assistant knows which page the visitor is on and tailors its answer accordingly.
- Open or gated flow: let visitors chat freely, or ask for name and email first, optionally only on specific landing pages.
- Smart handoff moments: the assistant recognises buying intent or a length threshold and then suggests getting in touch, instead of chatting indefinitely.
- Context-rich lead follow-up: when a visitor submits a request, a compact summary of the conversation is automatically attached, so a colleague never starts from zero.
- Answer quality feedback: visitors can rate each answer as useful or not, which feeds directly into how we tune the assistant.
- Admin dashboard: conversations, leads, handoffs and feedback are all tracked, with a built-in test button to check the connection to the AI provider.
- Fully no-code configurable: model, answer length, response speed, conversation duration, language choice, privacy text and retention periods are all adjustable through the settings screen, with no code changes required.
How it’s secured
A chat feature that talks directly to visitors and calls a paid AI service raises a fixed set of concerns. The same principle applies to any integration that talks to an external system: security belongs in the design from the start, not bolted on afterwards. On the main points:
- Access and permissions. Settings can only be changed by administrators, every action carries a security token so requests can be verified as coming from the website itself, and the AI key is never visible in the browser.
- Abuse and cost control. There’s a rate limit per visitor on messages, and a hard, server-enforced cap on the number of messages per conversation, so a single chat can’t run up unlimited AI costs.
- Privacy by design. No personal data is kept long-term in the browser, conversations aren’t stored at the AI provider itself, and what reaches a lead is a compact summary rather than a full transcript, with a configurable and automatically enforced retention period.
- Clean input and output. All visitor input and all AI responses are validated and sanitised before they reach any other part of the system (the website itself, email, database), and AI responses are limited to a restricted set of formatting rather than free HTML.
We’re deliberately sharing the approach here, not the technical implementation in detail, but it illustrates the principle: this kind of thing is built into the design upfront, not patched in afterwards.
In short
BlazeForce AI Chat is exactly what we deliver as BlazeForce when clients ask us for custom work: functionality that genuinely fits the practice, fully configurable without a developer involved, built on a reusable and maintained foundation, with security and privacy as a starting point rather than an afterthought.
Want a similar AI assistant on your own website or connected to your CRM? BlazeForce builds custom AI plugins that fit your systems, your security requirements and your tone of voice. Get in touch for a no-obligation conversation.
