Home / Work / Automated quoting

Quotes that draft
themselves overnight.

An Ontario industrial supplier was turning every inbound request into a quote by hand. We built an AI agent that reads the enquiry, prices it from their own rules, drafts the proposal and waits for a one-click human approval before it ever reaches a customer.

Manufacturing & distribution AI automation 6-week build Human-in-the-loop
Time to quote
−82%
from half a day to minutes
Win rate
+28%
faster replies close more
Saved per week
~11h
given back to the sales team
Logged in the CRM
100%
no quote slips through

The challenge

Every quote was a manual rebuild.

Requests arrived by email and web form in two languages, each one slightly different. A salesperson had to look up part numbers, apply tiered pricing and freight rules, copy it into a template, then re-key the whole thing into the CRM.

It took half a day to turn a busy inbox into sent quotes. Hot leads went cold while they waited, and rushed quotes carried pricing errors that ate into margin.

Our approach

Automate the draft, keep the judgment.

We didn't try to let an AI send prices unsupervised. We mapped the real quoting workflow, encoded their pricing logic as explicit rules, and used the language model only for the parts it's genuinely good at: reading messy requests and writing a clean proposal.

A working prototype ran on their last month of real enquiries within two weeks, so the team could judge it on their own cases before we hardened it for production.

What we built

From inbox to approved quote, on a rail.

01

Capture the request

Emails and web-form enquiries land in one place. The agent reads each one in English or French and pulls out parts, quantities and delivery details.

Intake
02

Price it by the rules

A deterministic rules engine applies their tiered pricing, customer-specific discounts and freight, so the numbers are theirs, not a model's guess.

Rules engine
03

Draft the proposal

The language model writes a clear, on-brand quote from the priced line items, in the customer's language, ready for a person to read.

LLM draft
04

One-click human approval

A salesperson reviews the draft, adjusts anything unusual, and approves. Nothing reaches a customer without that sign-off.

Review
05

Send and log to the CRM

On approval the quote is sent and written straight into the CRM with status and follow-up dates, so nothing is forgotten.

CRM sync

The outcome

The team reviews instead of retypes.

Most quotes are now drafted within minutes of a request arriving, even outside business hours. The salesperson's job shifted from data entry to a quick, confident review and send.

Faster, more consistent quotes lifted the win rate, pricing errors dropped to near zero, and every quote is now tracked in the CRM. The agent runs on their existing tools, and the team can see exactly what it did at each step.

Built with

OpenAI n8n PostgreSQL HubSpot Node.js Slack

Start here

What if your quotes wrote themselves?

Tell us the workflow that eats your week. A 30-minute call, in English or French, and an honest read on whether automation is the right fit.