The Website Assistant That Books the Meeting, Not Just the Chat

A B2B professional-services company in Cyprus was getting website enquiries around the clock but only answering them in business hours. Leads went cold overnight, follow-up was manual, and the contact form was a dead end that nobody chased. They wanted an assistant on the site, but they were nervous about it: their prospects are the kind of people who actually read the privacy policy, and a careless chatbot is an easy way to take on legal risk under the EU AI Act and GDPR. This is what we built for them, anonymized at their preference.

The situation

Every enquiry that arrived after hours sat unanswered until the next morning, and by then the visitor had often moved on. The team was doing the follow-up by hand, one message at a time, and the “someone will get back to you” promise was quietly costing them meetings. They did not just want a bot that chats. They wanted one that turns an interested visitor into a booked call, without a person having to be awake for it, and without creating a compliance problem in the process.

What we built

A bilingual (Greek and English) assistant that lives on their website as both a chat box and a voice conversation. It talks to visitors in plain language, works out what they need, and when they are ready it books them a discovery call on the spot, collecting a name, a contact detail, and a preferred time, instead of promising a callback. Every booking, and every “I want to talk to a human”, instantly alerts the team, and the visitor gets a branded confirmation email, a reminder before the call, and a follow-up if they do not show. A person confirms the slot and runs the call; the assistant does the catching, qualifying, and booking around the clock.

Why it stays compliant, from the first message

For this client the compliance was not an afterthought, it was the reason they had almost decided against a chatbot at all. So we designed it in from the first message rather than bolting it on later, across both bodies of law that apply, the EU AI Act and GDPR:

• It tells you it is an AI. The assistant discloses that it is an AI in its opening message, in the visitor’s language, and carries a permanent AI label, which is exactly what Article 50 of the EU AI Act asks a chatbot to do.
• It explains how your data is used, inside the conversation. A plain-language data-use notice is part of the chat itself (GDPR Article 13), not buried in a policy page nobody opens.
• Voice is treated as biometric data. Under GDPR Article 9 a voice recording is special-category data, so explicit consent is captured before anything is recorded.
• Nothing is kept forever. Conversations, transcripts, and recordings are deleted automatically after 30 days, with a second scheduled job as a safety net, and training on visitor data is switched off.
• There is evidence. A daily job writes a durable, personal-data-free record of each conversation, its consent basis, and what was processed, so the accountability trail outlives the 30-day deletion and can be handed to a regulator on request.

The result is an assistant the client can put in front of privacy-conscious prospects and defend on demand, rather than hope nobody asks. That is the difference between “we added a chatbot” and “we added a chatbot that survives an audit”.

What changed

• Enquiries are answered 24/7, in the visitor’s own language, instead of only in business hours.
• Calls get booked inside the conversation. The assistant collects the details, files the lead in the CRM, and alerts the team to confirm. No more “someone will get back to you”.
• The team hears about hot leads in seconds, through an instant alert the moment a call is booked or a human is asked for.
• Follow-up runs itself: a branded confirmation, a reminder before the call, and a win-back message if the visitor does not show.
• The assistant keeps itself current. When the client updates their pricing or services page, the assistant’s knowledge refreshes within a day, with no developer needed.
• Booking rate and lead quality: measuring. The assistant went live recently, so we are measuring these against real traffic and will report them once enough has accumulated. We do not publish numbers we have not measured.

How it works, in one paragraph

A conversational AI agent handles both the chat and the voice on the website. When a visitor is ready, the agent calls into our automation layer, which creates the lead and the tentative booking in the client’s CRM, sends the team alert, and triggers the customer emails. A set of scheduled jobs runs the compliance and self-improvement layer around it: nightly deletion of anything past its retention window, a daily consent and audit record, a weekly performance digest, and an automatic refresh that keeps the assistant’s answers in sync with the live site. We keep the specific products out of this write-up on purpose. The value is in the design and the controls, not the brand names on the toolbox.

Want a website assistant that books the meeting and stays audit-ready doing it? Talk to Dabster Labs about a compliant AI assistant for your site.

Related service: AI Chatbots