Automating quote generation in a B2B company: how to prepare quotes faster without losing margin

Take a hypothetical manufacturer of racking systems with 90 employees. A customer sends an enquiry with a drawing and a list of dimensions. The salesperson asks the technical department about the configuration, the warehouse about availability and the sales manager about the discount. The quote goes out after four days, although writing it took an hour.

In most companies the causes of delays and errors are similar. It is rarely about lazy salespeople – more often the data and decisions they need are scattered:

The measurable cost of manual quotes has two parts: people’s time and lost margin. You can calculate the first with a simple formula: number of quotes per month × average working time per quote × hourly cost.

Say there are 120 quotes a month and an hour and a half of work on each in total – by the salesperson, someone from the technical department and the approver. That is 180 hours a month. At a cost of PLN 80 an hour, this comes to PLN 14,400 a month.

You can check the second part on a sample. Take the last 30 quotes and compare the discounts given with the discount policy in force. If in some of them the discount exceeded the threshold without approval, calculate the difference in zloty – that is the margin automatic rules would have protected.

There is also a third cost: orders lost because of a slow response. It usually cannot be calculated reliably, so mention it to the management board, but do not add it to the total.

Quote automation is not a single tool but several connected elements. In a simple version they can all work within one CRM system; in a complex one, across several connected systems. The table below shows what each element does.

The hardest part of the project is not technical. In many companies only a few people know the pricing rules, and some of them have never been written down. A system cannot automate something nobody can describe.

Bring in the sales director, someone from finance and two experienced salespeople. Over a few meetings, agree answers to the questions in the list below. Record the result in a single document for the management board to approve.

This stage often pays off even before anything is implemented. Salespeople know what they can agree to on their own, and managers stop answering the same questions by email.

Not every company needs an elaborate quoting system. The choice depends on the number of quotes, the number of product variants and how complex the pricing is. The table below compares the three most common levels.

CPQ (Configure, Price, Quote) is a system that guides the salesperson through choosing the product’s parameters. It checks by itself which options go together and calculates the price. It mainly makes sense for configurable products, such as machines, made-to-measure furniture or installations.

In a company that works on Microsoft 365, a common mid-level solution is a Word template filled in by Power Automate with data from the CRM or a SharePoint list. Exceptions go for approval in Teams. Filling in the template does, however, require the Power Automate Premium plan.

AI is good with text. It can summarise a long request for quotation, extract a list of items and requirements, prepare a first draft of the scope description or tailor a product description to the customer’s industry. This shortens the salesperson’s work, but the output always needs reading.

An AI model should not calculate prices. It can get numbers wrong or pick a discount that sounds reasonable but breaks the policy. The rules engine calculates the prices, and AI uses its result.

Data protection also matters. Customer enquiries often contain confidential information, so use AI tools covered by a company agreement that do not use your data to train models. Employees’ personal accounts on free chat services should not have access to quotes.

Under the Polish Civil Code, an offer is binding on the company that made it. If the customer accepts it in time, a contract is concluded on its terms. That is why an automatically generated document must contain current prices and terms, not those from last quarter.

Every template should state how long the quote is valid. Without this, a quote sent by email is binding for the time in which a reply could normally be expected. How long that is can easily become a matter of dispute. If you do not want to be bound, call the document a commercial information sheet or a preliminary estimate, and state that it is not an offer.

For most contracts between businesses, acceptance by email or a click in a system is enough, as long as it is clear who gave it. A handwritten signature or a qualified electronic signature is only needed where the law or the contract itself requires written form. Agree the details with a lawyer when designing the template.

It is best to start with the one product line or customer type that generates the most quotes. The plan might look like this:

Without measurement, after the rollout you will only have impressions. Choose a few measures that can be calculated before and after the change. For most companies these are enough:

If a company sends a few quotes a month and each one is a tailor-made project, an elaborate system will not pay for itself. A shared template, a clause library and a tidy CRM are enough. The same applies when the CRM already has a good quoting module – it is better to configure it than to build a second one alongside it.

When automation does make sense, these are the mistakes that come up most often: