AI quotation assistant: from enquiry to draft quote

The problem

In most B2B businesses, a quotation starts as a loose message. It arrives by email, on WhatsApp or through a web form, with quantities, specifications and timing mixed into a few sentences. Someone reads it, opens the price list, applies the right rates and discounts, types the quote into the accounting system or a document, and sends it.

It is slow, and speed often decides who wins the order. It is also uneven: two people can quote the same enquiry differently. We have built assistants for this kind of work, and the pattern holds across distributors, manufacturers and service businesses. The rules exist somewhere. They just are not applied by software yet.

How it works

1. Receive the enquiry. The assistant monitors the channels you choose, such as a shared email inbox, web chat or WhatsApp.
2. Extract the details. A language model pulls out what is being asked for, quantities, timing and any special requirements, and asks the customer a short question if something essential is missing.
3. Price from your rules. Rates, discounts and terms come from your price list and commercial rules. The model applies them and never invents a price.
4. Draft the quote. A draft is prepared in your quotation template, or created in Tally, Zoho Books or your ERP or CRM.
5. Human approval. A salesperson reviews, edits if needed and approves. Standard, low-value quotes can be set to go out automatically if you choose.
6. Send and record. The quote is sent to the customer, and the enquiry, rules applied and approver are stored for follow-up.

Inputs, AI action and output

Input	What the AI does	Output
Customer enquiry in email, chat or form	Extracts requirements and quantities	Structured request
Price list and commercial rules	Applies rates, discounts and terms	Priced line items
Your quotation template	Fills it in and writes the cover note	Draft quotation
Salesperson decision	Records approval or changes	Sent quote and audit trail

Integrations and platform

We build on Google Cloud using Gemini models and agent frameworks, with a thin integration layer to your accounting software or CRM and the sheet or database that holds your price list. Your commercial team keeps maintaining prices where they already do, so the price list stays the single source of truth.

Guardrails and human review

• Prices only come from your price list. If an item is not found, the assistant says so rather than guessing.
• Drafts stay drafts until a person approves them.
• Every quote records the source message, the rules applied and who approved it.
• Customer data is handled with role-based access. Confirm retention rules with your adviser.

Data you need to start

• Your current price list and the discount or terms rules your team follows.
• Your usual quotation template.
• Twenty to fifty past enquiries with the quotes that were sent.
• Access to the system where quotes are stored, ideally a test version first.

A hypothetical walkthrough

This is a generic illustration, not a real customer. An email asks for a price on 500 units of an item and mentions a delivery date. The assistant extracts the quantity and date, finds the item in the price list, applies the quantity discount your rules allow and prepares the draft. The salesperson notices the cover note is a little formal, edits one line and approves. The quote goes out within minutes of the enquiry, and the full record is stored.

What changes for your team day to day

Salespeople stop retyping and start reviewing. Their time goes to the customer conversation and the difficult quotes. The owner of the price list becomes more important, because a clean list now drives every quote.

Rollout plan

We start by gathering your price list, rules and past quotes. We prototype on past enquiries and compare the draft quotes with the ones your team sent. Then we connect your system and run in shadow mode, where the assistant drafts and people decide. When the team trusts it, we open it to live channels. After go-live we maintain and improve it under an AMC.

Measuring whether it works

Agree a few simple measures before you start: time from enquiry to quote, share of drafts approved without edits, and the number of corrections per quote. For example, if your team currently takes half a day to respond, track how that changes over the first weeks. These are illustrative measures to track, not promised results.

Who it is for

Distributors, manufacturers and service businesses that quote often from a price list or rate card. It is a poor fit where every quote is a bespoke engineering estimate. For a deeper view, read our guide to AI quotation automation, or see matching enquiries to your product catalogue. Talk to us about your quoting process.