AI Chatbot and Lead Qualifier

01 The problem

Most website visitors leave without asking their question, and the ones who do often arrive out of hours. Generic chatbots either give canned answers or make things up, which costs trust. Sales teams were spending time on low-intent chats while high-intent buyers waited.

02 The approach

The assistant crawls a site’s pages, sitemaps, Notion and Google Docs, then answers through a retrieval pipeline: embeddings, vector search, a reranker and a streamed LLM response, with soft-404 and paywall detection keeping junk out. A 30 KB widget streams answers over server-sent events. Leads are captured mid-conversation and routed to Slack and email, team members can take over a live chat, and recurring unanswered questions become draft answers the owner approves in one click. Every workspace is isolated, and the platform ships with a WordPress plugin, billing and a super-admin console, backed by more than 2,600 automated tests.

03 Objectives

• Answer at least 80% of visitor questions without a person
• Reply in under 2 seconds, 24 hours a day
• Turn 5–8% of conversations into qualified leads
• Route hot leads to the team in real time through Slack and email
• Close knowledge gaps weekly from unanswered-question reports

04 Key features

• Retrieval-augmented answers from your own content
• URL, sitemap, Notion and Google Docs ingestion
• Lead capture, scoring and routing
• Live human takeover
• Self-improving answer suggestions
• 30 KB embeddable widget and WordPress plugin
• Multi-tenant workspaces and billing
• Analytics on answers, leads and conversion

05 Links and services

• AI Assistants & Knowledge Systems
• Digital Systems & Integrations