Short answer: quote and proposal automation is the work of turning a qualified inquiry into a signed contract without anyone rebuilding a price in a spreadsheet or retyping agreed terms into a document. It has four stages, and most teams automate the last one first.
The last stage is the document. It is the visible one, so it gets the software budget. A proposal tool arrives, the templates look better, and the quote still takes six days, because the delay was never in the typing.
Where a quote actually gets stuck
Take your last ten quotes. Record two timestamps for each one: when the qualified requirement landed, and when the buyer received a document they could sign. Then split the elapsed time into three buckets.
Internal wait is time the request spent with your own people. Someone had to build a price, check a margin, or approve a discount. Information wait is time spent asking the buyer for facts you needed before you could price at all: volumes, site conditions, the systems you would have to connect to, the legal entity on the contract. Calendar wait is time the request spent waiting for a meeting that had to happen before anyone would commit to a number.
Read the timestamps rather than estimating from memory. Estimates flatter the internal bucket and hide the other two. The quote was not slow to write. It was slow to become writable, and that is a different problem. A better template does not touch it.
Stage one: capture the requirement once
A quote needs a small set of facts, and the same facts every time. Most companies collect them three times: once in the inbound form, once in the discovery call, and once again when the person building the price discovers that the first two versions disagree.
Fix the shape of the handover before you automate anything downstream. Name the fields a price cannot be built without, make the qualification step responsible for filling them, and refuse to start pricing on a record that is missing one. That refusal is the automation. It feels like friction and it removes days.
Prefill whatever public data already answers. In Estonia the e-Business Register returns the legal name, registry code and board members for any company from a single identifier, so asking a buyer to type their own registry details into your form is a question you can delete. Every deleted question is a wait you no longer have to chase.
If the inbound step itself is the bottleneck, that is the stage before this one, and we wrote it up separately in sales process automation. Fix the inquiry response first if inquiries are sitting unanswered overnight. Quoting faster on a lead you answered three days late changes nothing.
Stage two: price it from a rule, not from memory
This is the stage that decides whether the rest can be automated at all, and it is the one companies skip.
Write down what you sell as named items. Give each one a rule: the base number, the unit it scales with, and the specific conditions that change it. Distance, volume, urgency, site access, integration count, support hours, whatever genuinely moves your cost. Then write down the discount bands and who may approve each one.
Two things happen when that document exists. Quotes stop disagreeing with each other, which is the quiet cost nobody measures until a buyer compares two of your prices. And the document layer becomes a render step instead of a writing task, because there is now something to render.
Until the rule exists, every quote is a fresh act of judgement, and no tool can assemble a fresh act of judgement. Teams that buy quoting software before writing the price list end up with an expensive template editor. We describe the same failure from the buyer side in how to compare AI automation quotes: a proposal that cannot show its own arithmetic is hiding it.
Keep the judgement calls out of the rule on purpose. Anything you cannot state as a condition is a human decision, and marking it as one is a design choice rather than a gap.
Stage three: assemble the document
Now the render step. Pull the priced items, the buyer entity, the scope text for each item you sell, and the standard terms. Produce a document. Nothing here should require a person to retype a number that already exists in a system.
Two rules keep this stage from rotting. First, one source for each fact. If the price of an item lives both in the price list and in a paragraph of template prose, they will drift, and the paragraph will win in front of a buyer. Second, version the terms. When your standard terms change, you need to know which quotes went out under which version, and the only reliable way is to record the version on the document itself.
Language models help here, and the useful job is narrow. They are good at turning a structured scope into readable prose for a specific buyer, and at drafting the cover summary. They are not good at deciding a price, and a model that composes a number rather than reading one is a defect, not a feature. Give the model the priced record and let it write around the numbers, never in place of them.
If you are choosing between an off the shelf quoting product and something built to your own rules, that trade is the subject of build versus buy for mid-market software. Buy the standard part. Commission only the part that encodes how you actually price.
Stage four: approval and signature
Approval is routing. A discount inside the standard band goes straight out. A discount outside it goes to the named person who owns that band, with a due time and an escalation if the time passes. Most approval delay is not disagreement, it is a request sitting unseen in an inbox with no clock on it.
Signature is a legal step rather than a formatting step, and it is worth knowing the ground you stand on. In the European Union the eIDAS Regulation gives a qualified electronic signature the same legal effect as a handwritten signature, and states that a signature may not be denied legal effect only because it is electronic. In Estonia the national signing tooling is documented at id.ee. In the United States the ESIGN Act of 2000 takes the same position for interstate commerce, and the UNCITRAL Model Law on Electronic Signatures is the template many other jurisdictions followed. The United Kingdom Law Commission reached a comparable conclusion for documents executed under English law in its work on electronic execution of documents.
The caveat that matters in practice: specific contract types still carry national form requirements, and sector rules can add more. Confirm the ones that govern your own contracts before you remove the last paper step. Being right about the general rule and wrong about your one regulated document is an expensive way to be correct.
One more boundary. A quote record holds personal data: names, direct contact details, sometimes identity details for the signatory. The GDPR principle of data minimisation asks you to hold what the purpose needs and no more, which is a useful editing rule for intake forms as well as a legal duty. If a field never changes a price and never reaches the contract, stop collecting it.
What stays with a person
Scope. Any discount outside the published band. Any commitment on a delivery date that the delivery team has not confirmed. Every non standard legal term. Anything where a wrong answer costs you money you cannot recover, which in quoting means anything that ends up binding.
Automation may gather facts, apply a published rule, assemble the document, route it for approval and chase the signature on a schedule. It may not invent scope, and it may not agree terms nobody has read. Draw that line in writing before the build, because it is the line that decides how much of the flow you can safely leave alone.
A four week sequence that does not stall
Week one, measure. Ten quotes, two timestamps each, three buckets. Do not change anything yet. You are looking for which bucket is largest, and the answer is regularly not the one the team predicts.
Week two, write the price list. Named items, a rule each, the conditions that change the number, and the discount bands with owners. This week needs no software at all and it is the week that makes every later week possible.
Week three, fix the largest bucket, and only that one. If it is information wait, cut intake fields and prefill from the register. If it is internal wait, give each approval a named owner, a due time and a timed escalation. If it is calendar wait, work out which meeting can be replaced by a written answer.
Week four, render and route. Now assemble the document from the priced record, version the terms, and put the signature step on the same clock as everything else. Then measure the same ten numbers on the next cohort of quotes. If you want the first week measured on your own deals rather than estimated, our free AI audit for a business is the version of week one that we run for you.
Adjacent problems this is not
If your question is about answering the inquiry in the first place and deciding whether it is worth quoting, that is sales process automation and it sits before this flow. If it is about what happens after the signature, when a new client has to become a working account, that is customer onboarding automation. If it is about the supplier invoices arriving at your own accounts payable desk, that is invoice processing automation, and it is a buyer side flow rather than a seller side one. If you bill for time rather than for scope, the pricing input is different again and we covered it in automation for professional services and billable hours.
Frequently asked questions
What is quote and proposal automation?
It is the work of turning a qualified inquiry into a signed contract without anyone rebuilding a price in a spreadsheet or retyping agreed terms into a document. It covers four stages: capturing the requirement, pricing it, assembling the document, and getting it approved and signed. The document stage is the visible one, so it usually gets automated first, even though the delay rarely lives there.
Why is our quote slow when writing it takes twenty minutes?
Because the writing is not the wait. Measure the elapsed time on your last ten quotes and split it into three buckets: time spent with your own people building or approving a price, time spent asking the buyer for facts you needed before you could price, and time spent waiting for a meeting. The second and third buckets are usually larger than the first, and a nicer template touches neither.
What should we automate first in the quoting process?
The price list, not the template. Write down what you sell as named items with a rule for each one, including the conditions that change the number. Until that exists, every quote is a fresh act of judgement and no tool can assemble it. Once it exists, the document becomes a render step rather than a writing task.
Is an electronically signed quote legally binding?
In the EU, eIDAS gives a qualified electronic signature the same legal effect as a handwritten one, and no signature may be refused legal effect only because it is electronic. The United States ESIGN Act and the UNCITRAL model law take a similar position. National form requirements still apply to specific contract types, so check the ones that govern your sector before you remove paper.
What has to stay with a person when quotes are automated?
Scope, any discount outside the standard band, any commitment on a delivery date the team has not confirmed, and every non standard legal term. Automation may gather facts, apply a published rule, assemble the document, route it for approval and chase the signature. It may not invent scope or agree terms that nobody has reviewed.
How much does quote and proposal automation cost?
It depends on how many items you sell, how many conditions move the price, and how much your current systems already hold. Our model starts with a free 30 minute audit that maps the flow and puts the wait in numbers, and build and run is quoted after the audit.
If you want those three wait numbers measured on your own last ten quotes instead of estimated, that is what the free audit is for. Tell us how a request for a price reaches you today and we will map where the days are going.