WhatsApp chatbot for lead qualification

The problem

In sales over WhatsApp, what kills a deal is not a poor answer — it is the wait. The customer sends a message, sits in silence, and by the time someone replies they are already talking to a competitor. Speed of first contact is a conversion factor in its own right, regardless of what gets said afterwards.

What it does

It handles the first message, answers the questions that keep repeating, and asks what is needed to qualify the lead before passing the conversation on. The conversation reads as human — no “press 1 for sales” — because support that announces itself as a robot loses the person just as fast as silence would.

And it feeds the CRM on its own: whatever the conversation surfaces becomes a record, with nobody transcribing it later. That is the part usually abandoned in practice — the salesperson does reply quickly, but never logs it, and the information dies on their phone.

The goal was never to automate support end to end. It was to make sure that when someone from the team joins the conversation, the context is already gathered and the lead is already filtered.

The result

The company’s conversion rate went from 2% to 3.6% — in a high-average-ticket business, where every percentage point is worth a lot. More than three thousand customers went through the bot before any conversation reached a person.

Both figures are the client’s, not from a test of mine — which is why the card says “from the project” rather than “measured”. They are verifiable with the people running it.

It was built in a pair, at Dark Marlin. I split the implementation with another person; this isn’t a project I wrote alone and I won’t present it as one. The code isn’t public.