Business impact

<aside> 💡

The problem: Sales managers were spending [X hours/week] manually digging up objection responses and relevant case studies for every client conversation, and commercial proposals sat waiting on manual back-and-forth before they could go out.

What it looked like before: Winning arguments and proven case studies existed only in individual managers' heads or in a spreadsheet nobody actively searched — so proposal quality depended on who happened to be writing it, and prepping for a call meant digging through CRM notes by hand.

What we built: An AI Sales Intelligence Platform that surfaces the most relevant case studies and objection-handling arguments automatically, generates a tailored commercial proposal with manager approval before it reaches the client, and prepares a ready call script for any lead in seconds.

Result: What used to mean digging through CRM notes and relying on memory now takes seconds — the system pulls the most relevant case study and drafts a call script automatically. Every proposal draws on the company's full win history instead of one manager's memory, so quality no longer depends on who happens to be writing it.

</aside>

AI system that turns a company's own sales history (past objections, won deals) into a searchable knowledge base, then uses it to generate personalized commercial proposals and call scripts on demand for a sales team using Zoho CRM.

Project details

Context:

• Business/industry: Sales / B2B CRM operations
• Current process: Managers write commercial proposals and handle client objections from memory or scattered notes; there's no shared, searchable record of "what worked before" for a similar client or a similar objection.

Problem:

• Winning cases and proven objection-handling scripts exist, but only in a manager's head or in a spreadsheet no one searches.
• Proposal writing takes time and quality depends on which manager is writing it.
• No fast way to prep for a call with fresh, CRM-sourced context.

Solution

<aside> 🧠

A shared vector knowledge base (Qdrant) of sales objections and case studies, queried by semantic similarity — not keyword search — to ground two automations: an AI-generated commercial proposal pipeline with manager approval, and a lead call-script generator. A separate objection-handling bot is scaffolded but not yet wired up.

Tools used: n8n, Qdrant (vector database), Google Gemini Embeddings, Claude Sonnet 4.6 / Llama 3.3 70B (via OpenRouter), Zoho CRM, Google Sheets, Google Docs, Gmail, Telegram.

</aside>

Solution logic:

1. Knowledge base loader: pulls sales objections and case studies from Google Sheets, prepares each row into a structured text block, embeds it with Google Gemini, and stores vector + full payload (strategy, response template, case details) in a single Qdrant collection.