Automating quotes and proposals: what we build and what stays manual

Most of a quote is assembly, not judgment. What AI and automation take over in quoting, where a person still decides, and how to keep quotes consistent.

3 minread 655words last updated

The short answer

Quoting feels like skilled work, and part of it is. Most of it is not: reading the request, finding the right sections and prices, assembling a document in the right format, sending it, and following up three days later. That part is assembly, it is the same every time, and it is where quotes get slow and inconsistent. Automation and AI take the assembly. A person keeps the judgment and approves every quote before it goes out.

Take the process apart

StepTodayRule or judgmentWho does it after
Request arrivesSits in an inbox until someone reads itRuleAutomation: captured on arrival
Understand what is askedSomeone reads and interpretsJudgmentAI: extracts scope, quantities, dates, constraints
Decide what to offerSomeone thinksJudgmentPerson, with the extracted request in front of them
Assemble the quoteSomeone copies from the last oneRuleAutomation: sections and prices from your building blocks
Write the personal partsSomeone typesJudgmentAI drafts; person edits
Check and sendSomeone forgets for two daysRulePerson approves; automation sends and logs
Follow upSomeone remembers, or does notRuleAutomation reminds; AI drafts; person approves

What we build

  1. Capture. Requests from email, forms and notes land in one place, with the original attached.
  2. Extract. AI reads the request and produces a structured summary: what, how much, when, any constraints or unusual asks. Flags anything unclear for a question back to the customer.
  3. Assemble. From the summary and your rules, automation selects the sections and fills in the numbers from your price list. Anything outside the rules is flagged, not guessed.
  4. Draft the personal parts. A short introduction and the specific notes, drafted by AI in your tone from the request and previous quotes.
  5. Approve. The salesperson reviews the whole quote, adjusts judgment items, and sends with one click. Nothing leaves unseen.
  6. Follow up. After the agreed interval, a reminder; a drafted follow-up referencing the quote; a person approves. Repeat until answered.

What stays manual

  • Pricing decisions outside the price list and rules: discounts, unusual volumes, strategic accounts.
  • Scope judgement on anything the building blocks do not cover.
  • The relationship: the call, the meeting, the negotiation.
  • Genuinely bespoke work, where the system still helps by reading the request and preparing the skeleton.

What this means for you

If quotes take days to go out and each one is written from the last one, the assembly is eating your sales team’s time and making your offers inconsistent. Write the building blocks, automate the capture, extraction, assembly and follow-up, and keep a person on the decisions and the send button. Quotes go out the same day, say the same things, and get followed up every time.

Written by the CivSec S.M.A.R.T team

We build and run websites, software and AI systems for businesses. We write about what we see in that work, in plain language, and we update articles when things change.

Last checked . Spotted something outdated? Tell us.

Frequently asked questions

Can the AI set prices?

It can apply your price list and your rules, which is automation. It should not invent prices or discounts, which is judgment. We build systems that fill in the numbers from your rules and flag anything outside them for a person. A model guessing a price is how margin disappears.

Our quotes are all different. Can this still work?

They are more similar than they feel. A quote is the same sections in different combinations with different numbers. The first step is to write those sections once, well. Genuinely bespoke quotes stay manual, and the system still helps by reading the request and preparing the skeleton.

Where does the request come from?

Email, a website form, a phone call written up, a chat. The system reads whatever arrives, extracts what is being asked for, and matches it to your building blocks. A form with structured fields makes this easier; free text still works with an extraction step.