Start quote automation with a simple question: which steps genuinely repeat? If every enquiry requires a new design, a site visit and a technical decision, a basic calculator will not cover the entire process. You can still organise information, flag missing details and prepare a draft response. The first scope does not have to automate everything.
With made-to-measure products, the same questions often return: width, height, material, finish and delivery. Building my own brand, CopperAndCraft, taught me that identifying these details correctly matters just as much as doing the arithmetic. This guide explains how to prepare a pilot. It does not report results from automated enquiry handling implemented for clients.
Map the journey of one enquiry
Select a few typical messages and retrace your work. What do you read from the email? What do you copy into a spreadsheet? Which details come from a price list, and which require another question? Note the points where someone needs to make a judgement. This produces a simple process description that can later be tested.
Begin with one product type or service. Avoid combining every market, exception and pricing method in the first version. A limited scope makes it easier to check whether the solution behaves correctly. Adding a second product should follow an assessment of the first, rather than an assumption that all cases are similar.
Separate understanding a message from calculating a price
AI can help identify parameters in a message or draft a reply in the customer’s language. Explicit rules should determine the price. If pricing depends on area, the system needs an approved unit, conversion method, minimum order and rounding rules. A special finish needs its own surcharge rule.
An illustrative calculation might be: preparation cost + area × rate + finish surcharge. This demonstrates a pricing structure only. Shipping, taxes, discounts and currencies need separate rules suited to the business. They should not be guessed from similar orders.
A useful distinction: a completed draft is not an approved proposal. Product details, price, lead time and delivery terms still need to be checked before sending.
Decide what happens when information is missing
A useful automation must be able to stop. A missing height does not justify assuming a standard dimension. A missing delivery country does not mean domestic shipping. Conflicting dimensions in the subject line and message need clarification. The correct output may be a list of missing fields or a draft question, rather than a price ready to send.
Define the required fields: product, material, dimensions and units, quantity, finish and information needed to assess delivery. Specify allowed values for each field. Also distinguish between gaps that prevent any calculation and those that allow an indicative product price without shipping.
Check the quality of your sources
A disorganised price list remains a problem after AI is connected. If rates exist in several files, choose an authoritative source and assign responsibility for updates. Products need recognisable identifiers. Commercial names alone may not identify a variant unambiguously.
The same applies to descriptions. Material, installation and limitation information should be consistent. Preparing the product catalogue can be a useful step before automating replies. When adding a sales channel, also consider the practical issues covered in the Etsy to Shopify migration guide.
Design a focused pilot
A pilot needs a defined input and output. For example, enquiries about one panel type become structured fields, a calculator produces a product estimate, and the user receives a draft response. The system does not send proposals independently yet. A person compares the output with a manually prepared quote.
Before starting, agree which applications are involved, which permissions are necessary and where information may be processed. Use examples without identifiable customer details for initial trials. Access should match the task. Testing a handful of sample enquiries does not automatically require access to an entire inbox.
Test exceptions, not only complete messages
- Dimensions in different units or with decimal commas.
- A missing dimension or quantity.
- A material or variant absent from the price list.
- Conflicting information or a revised dimension later in the conversation.
- A duplicate message that should not create a second enquiry.
- An unavailable pricing source or a failed connection.
Each case needs an expected result: a correct calculation, a clarification question or a clear stop message. A reply appearing on screen is not enough for acceptance. You should also be able to explain why a particular amount was calculated.
Compare costs and actual work before and after
Measure several real cases before claiming a time saving. Include reading, correcting information, checking the calculation and preparing the reply. After the pilot starts, measure the same end-to-end task. Generating text quickly may achieve little if reviewing it takes as long as the previous process.
Include implementation, tool subscriptions, model usage, monitoring and later pricing changes in the costs. Agree how the automation can be stopped and what the manual fallback looks like. Ongoing support may be a separate service, but its scope should reflect real responsibilities rather than an AI label.
Explore an example, then discuss your workflow
The service page includes an interactive quote preparation demonstration. It uses predefined enquiries and illustrative rates. It is not connected to an AI model or an email account. The example shows the difference between complete information and a case that requires a follow-up question.
If a similar task repeats in your business, start with a process description and sample data. An enquiry and quote automation pilot provides a way to agree a limited scope, acceptance criteria and costs before expanding the solution.