How to automate quote generation with AI

By Péter Ujvári · May 3, 2026 · 7 min read
TL;DR A typical B2B quote takes 30–90 minutes of a salesperson's time. With AI you can bring that down to 5–10 minutes while improving consistency and traceability. The key is a well-structured pricing source + LLM + human approval. Such a system ships in 3–6 weeks at €3,500–7,500.

Why is quoting slow today?

Because every salesperson does the same thing for every quote: find the price list, open a template, copy, format, double-check. About 80% of that knowledge is reusable — exactly what LLMs are good at.

5 steps to a working system

  1. Map data sources. List your price lists (Excel, ERP), templates (Word, PDF) and past quotes (CRM, email). These become the AI's knowledge base.
  2. Lock template + tone of voice. A unified template plus 3–5 example quotes makes the LLM produce consistent output.
  3. LLM + function calling. The model uses function calling for the pricing math (zero numerical errors) and freely generates the prose.
  4. Human-in-the-loop UI. Salespeople see the generated quote in a simple web UI, edit and approve it. Nothing leaves without a human.
  5. Measure. Two KPIs: time saved per quote and change in win rate. Both are visible within 2 weeks.

Pitfalls

The most common mistake is letting the LLM do the math — don't. Prices, discounts and VAT must come from code; only the prose and structure come from the LLM. That zeroes out numerical errors.

Expected outcome

Typical measurement at a 6-person sales team: 70% turnaround reduction, +20% quote volume (with the same headcount), and a small (5–10%) win-rate uplift because faster responses win more business.

Want me to map your current process? Email me for a free 30-minute call.