Skip to content
caadoo
de

How to automate quoting, and where a person still signs off

Automated quoting rarely means one tool that does everything. It means price rules and templates in the CRM or ERP you already run, an AI draft where requests arrive as free text, and a named person who approves anything unusual before it reaches the customer. Here is how that splits across the quoting process.

By T. Krause11 min read

A long-standing customer emails on a Monday asking for a price on 200 units of a part they have bought for years. The quote goes out on Thursday. Nothing about it was hard. The rep who knows that customer's discount was at a trade fair, and nobody else wanted to guess.

That delay is what most companies mean when they say they want to automate quoting. This piece is for owners and sales managers in companies of roughly 20 to 250 people. It walks through the quoting process one step at a time and says, for each step, what software can do on its own, what still needs a person, and where the sign-off sits. Every tool named here is described only by what its vendor's own documentation says it does.

Photo: Unsplash

What automated quoting covers

Vendors use the phrase for very different products. It helps to separate three kinds of automation, because they fail in different ways.

Price rules. Customer-specific prices and discounts, a validity date, a threshold above which a manager has to approve. Rules are boring and predictable, which is exactly what you want near a price.

Templates and master data. Your letterhead, item list, standard terms and text blocks live in the CRM or ERP once. A quote is assembled from them instead of being copied from the last similar quote, old mistakes included.

AI drafting. A language model reads a request written in free text, proposes line items, writes the follow-up question or the cover email. It saves typing and gets things wrong now and then, so its output is a draft and stays one.

If your prices still sit in a spreadsheet only one person understands, start with the first two. A model that drafts quotes against unreliable prices just produces wrong quotes faster.

The quoting process, step by step

StepWhat can run on its ownWhat a person keepsSource
Request intakeWatch a mailbox, match the customer, check stock, create the quoteConfirm the incoming requestMicrosoft, Sales Order Agent
Clarifying requirementsEmail the customer about missing detailsReview each outgoing email, handle custom specsMicrosoft, Sales Order Agent
PricingSpecial prices, line and invoice discounts filled in from master dataPricing anything non-standardMicrosoft, sales quotes
Text and line itemsBranded templates, saved items, alternative and optional linesChecking the description fits the jobLexware Office, quotes (German)
Drafting the wordingFirst draft from notes, outlines or reference filesChecking facts, numbers and namesMicrosoft, Copilot in Word
ApprovalRoute quotes above a discount threshold for approvalApproving or rejectingHubSpot, quote approvals
SendingPDF attachment with a prefilled email, online acceptance with a notificationSending once approvedMicrosoft, Lexware Office
Follow-upValidity date set automatically from the creation dateThe call, and any change to the priceMicrosoft, sales quotes
Record keepingArchive a version when a quote becomes an orderDeciding how long to keep itMicrosoft, archiving documents

These products appear because their vendors document the feature publicly, not because they are the only ones. Most CRM and ERP systems have something similar. Check what your own licence includes before you buy anything new.

Intake and clarification: where AI earns its keep

Requests arrive as emails, web forms, phone notes and PDF tenders. Turning that free text into line items is the part of quoting where a language model is most useful, and it is also where most of the waiting happens.

Microsoft's Sales Order Agent for Business Central is a concrete example. According to Microsoft, it monitors a designated mailbox, verifies the customer, checks item availability, creates a sales quote and prepares a reply with the quote as a PDF. Until an order exists, it handles the email exchange with the customer about missing details. A person reviews incoming requests and outgoing emails in a task pane before anything moves on.

Microsoft is also open about the limits. Its page lists cases where quantities need correcting because units of measure differ, and notes that in some local versions a search term is missed when it appears inside a longer word. With a German item catalogue full of compound nouns, plan for it. If you sort a shared mailbox this way, the model sorts and drafts and a person sends, as described in shared inbox triage with AI.

Pricing stays with the system, not the model

A language model should never calculate a price. Prices come from master data, and for standard goods that part has been automatable for years without any AI. In Business Central, a quote line updates itself when a special price or line discount is set up for that customer or item and the condition is met. Lexware Office says its quotes pick up the conditions stored for a customer or item automatically.

If your price logic lives in a workbook with hidden tabs, fix that before anything else. Replacing Excel with an internal tool describes how to move the prices out first and leave the calculation alone until the base is stable.

Before you paste customer data into a public chatbot to draft the wording, read ChatGPT and the GDPR. Copilot in Word can draft from notes, outlines or reference files, and Microsoft's own help page tells you to verify facts, numbers, names and citations and says you are responsible for what stays in the document. To draft a single quote that way today, with a prompt, a worked example and a checklist, see how to create a quote with AI.

Where a person signs off

The sign-off is a rule about which quotes a person has to see. Everything else can go out after a quick check by whoever prepared it.

HubSpot's documentation shows what such a rule looks like in practice. Quote approvals can be required above a set discount amount, overall or for a single line item. Enterprise accounts can also trigger approvals on the quote amount, the SKU, billing frequency or payment terms.

For a company of 50 people, I would start with four triggers: a new customer, a discount above your limit, a custom build, and a total above a set amount. A quote that hits none of them goes out after the person who prepared it has read the PDF. A quote that hits one goes to a named manager first. Human in the loop explains how to place that line by asking whether a mistake can be undone and what it costs.

Follow-up without a spreadsheet of reminders

A quote with no reply tends to sit until someone remembers it. Business Central can fill the quote's valid-until date automatically, a set number of days after the quote is created. That date is the hook for a reminder, say a week before it runs out.

Lexware Office lets customers accept a quote online and sends you a notification when they do. A model can draft the follow-up email. Whether to chase, wait or offer a better price is a decision for someone who knows the customer. How to automate quote follow-up has a sample cadence and the cases where a call does more than another email.

If you sell under German law

For a company selling from Germany, an automated quote is as binding as a hand-typed one. Under Section 145 of the German Civil Code (BGB), whoever offers to enter into a contract is bound by the offer unless they excluded that. If you set a deadline, Section 148 limits acceptance to that deadline. Without one, Section 147(2) allows acceptance until the point at which you could expect a reply under normal circumstances. If a wrong price in your master data reaches every quote for a month, you are bound by every one of those offers.

Copies of sent business letters must be kept for six years under Section 257 of the Commercial Code (HGB) and Section 147 of the Fiscal Code (AO), counted from the end of the year they were sent. Under the fifth sentence of section 147(3), the Fiscal Code period does not run out as long as the letter matters for a tax whose assessment period is still open. Whether a quote that never became an order counts is a question for your tax adviser. Either way, check what your system does with the quote once it converts. Microsoft's documentation says converting a quote in Business Central removes it from the database, and that automatic archiving can save a version when that happens. This is a summary of the statutes, not legal advice.

An invented example

The company below does not exist. Picture a 35-person distributor of fasteners and fixings that gets most of its requests by email from trade customers who already have an account.

Requests from known customers for stock items run straight through. A program reads the mailbox, creates the quote in the ERP with the customer's agreed prices, and a sales assistant opens the PDF, reads it and sends it. Requests from new customers and any quote with a discount above the agreed limit go to the sales manager's approval list. Requests for anything not in the item list are left alone and go to a person as before. A model only summarises them into a list of open questions.

No step in this design sends a quote that nobody has looked at. You can relax that later, once the drafts have been right for long enough that the checking feels like a formality.

Where to start

Count before you build. For one week, note how many requests come in, how many are standard, and how long each waits before the quote goes out. What to automate first shows how to do that with a tally sheet, and measure before you build explains why the baseline matters when you later ask whether it paid off. The count tells you whether your bottleneck is the inbox, the pricing or the approval, and that is where the first piece of automation goes.

caadoo is a monthly subscription, and setting up this kind of quoting pipeline is typical work in it. I go through real requests with your sales team, get the prices into one place, agree the approval triggers with you, and then build the automation so a person sees every quote before it leaves for as long as you want that. How it works explains the subscription. If quotes are piling up in your inbox right now, tell me which ones.

Frequently asked questions

What is quote automation?

Quote automation means software assembles a quote from your customer data, item list, prices and templates instead of a person typing it. Rules handle discounts, validity dates and approval thresholds. An AI model can add a first draft when requests arrive as free text. A person still approves unusual quotes before they go out.

What does automated quoting software do?

It creates quotes from master data, applies customer-specific prices and discounts, routes quotes above set limits for approval, sends them as PDFs or for online acceptance, and converts accepted quotes into orders or invoices. Many CRM and ERP systems already include this, so check yours before buying a separate tool.

Can AI write quotes?

It can draft them. A language model reads a free-text request, proposes line items and writes the cover email. Prices should come from your system, not the model, and Microsoft itself tells Copilot users to verify facts, numbers and names in every draft.

How do you automate quote follow-up?

Set a validity date on every quote automatically and create a reminder from it, for example a week before it expires. Some tools notify you when a customer accepts online. The follow-up call and any decision about a better price stay with someone who knows the customer.

Is a quote legally binding?

Under German law, yes, unless you excluded it. Section 145 of the Civil Code binds whoever makes an offer, and Section 148 limits acceptance to any deadline you set. That holds for quotes created and sent by software too. Other countries have their own rules, so check the law that governs your contracts.

Topics

  • Quote automation
  • Sales
  • Automation
  • AI
  • Mittelstand
All articles

Keep reading

  • 19 min read

    Microsoft 365 Copilot rollout checklist: 8 steps

    Copilot can find anything a person is allowed to open in Microsoft 365, and in most tenants nobody knows exactly what that is. This checklist puts the work before the first licence in order: permissions, data protection, works council, training, policy, pilot and measurement.

  • 12 min read

    Process documentation template: how to document a process

    Process documentation that people use fits on one page and describes the work as it is done today. Here is a template you can copy, seven steps for documenting a business process, a worked example from a 120-person manufacturer and the mistakes that leave these documents unread.

  • 11 min read

    Is ChatGPT GDPR compliant for business? A checklist

    Whether ChatGPT is GDPR compliant depends less on OpenAI than on the plan you buy and the paperwork your company does around it. This piece covers which plans come with a data processing agreement, training, data location and the duties that stay with you.

End of call

You can hang up now.

Or book 15 minutes. I'll ask you the questions I'd ask on the phone, and tell you at the end whether it fits. If it doesn't, we've both lost a quarter of an hour instead of a month.

Or call. Cold, if you like. My AI assistant picks up and I call you back. +49 5136 9414999

Analytics

With your consent I use Google Analytics to see which pages get read. Without it, no analytics code loads. Your answer is stored locally in your browser.

Privacy notice