Levelstate Systems supplies engineering goods. They wanted every website and WhatsApp visitor to turn into a clear lead. 24vertex built AI chat agents that run inside a self-hosted Chatwoot inbox. The project went live in 1 month.

We connected a website chat widget and a WhatsApp channel to one place so the sales team could watch every conversation. The agents use retrieval augmented generation, or RAG, to answer product questions from approved content. When a buyer asks for pricing or a quote, the agent collects contact details and passes the chat to a human. The team sees the full context and replies faster.

The challenge

Levelstate lost incoming leads. Their tawk.to bot handled FAQs well, but it failed to capture names, emails, or phone numbers. Visitors left without leaving details, and sales had no clear path to follow up. Conversations sat in different places, which made tracking hard and raised the chance of human error.

The 24vertex architecture

We deployed a self-hosted Chatwoot inbox and created two channels, one for the website widget and one for WhatsApp. Every chat now lands in a single queue that the Levelstate team can monitor in real time.

We built RAG-based AI agents that read Levelstate-approved product content. The agents find answers in that content and reply in plain language. When a buyer shows intent, the agent asks for contact details, records them, and signals a handoff. A sales representative picks up the thread inside Chatwoot without switching tools.

We set up a self-hosted Supabase instance as a vector database. It stores embedded versions of Levelstate documents so the agent can retrieve the right facts quickly. Chatwoot provides the chat interface and the WhatsApp connection, while Supabase powers accurate answers. Together they create a clear path from question to qualified lead.

Outcomes

Levelstate now captures lead details that the old bot missed. The sales team watches and replies to website and WhatsApp conversations from one platform. The AI handles complex engineering questions and routes handoffs to the right person. The system builds a growing database of everyone who engaged, which strengthens follow-up and forecasting. We delivered the project in 1 month, and the internal team manages the combined inbox day to day. The efficiency and data quality support 24vertex's 300% ROI standard for automation work.