Deep DiveNo. 06
Demand is growing. Can your quoting keep up?
Less time spent preparing quotes, more capacity to respond to customers, and a system that improves as the team documents how the business works
Audio could not be loaded. Download the MP3 or read the issue below.
Three in four quotes produced by the system need no corrections. Based on the reviews available, we estimate that 77% have the right products, quantities, prices and scope. A member of the sales team reviews and approves a draft rather than building it from scratch.

At a fire protection company receiving an average of 800 quote requests a month, reducing the work involved in each quote allows the sales team to respond to more requests.
For businesses with high sales volumes and tight margins, handling demand efficiently matters just as much as generating it. Every quote takes time to prepare, whether or not it leads to a sale. Across hundreds of requests, that time adds up and eats into margins.
In this case, the team could only keep up with demand by putting in the hours. Every request involves interpreting information, selecting products and working out prices. If all of that is manual, handling more opportunities means hiring more people or making customers wait longer.
As we saw in our case studies on customs declarations and RFPs, the constraint is how much preparation the team can handle. The opportunity is to process that volume while freeing the sales team to review quotes, look after customers and close deals.
01
From request to quote, without starting from scratch
The system handles requests through WhatsApp. A salesperson logs requests from other channels in the CRM, the customer relationship management system, so they follow the same workflow.
Customers simply explain what they need on WhatsApp. AI interprets the message and asks follow-up questions to gather the information needed to prepare a quote.
As we saw in the RFP case, AI makes automation more flexible: it makes sense of requests however they are worded and turns them into structured data that other software can use.
Next, the system needs to select the products, work out the price and check which terms apply. For this project, the team’s knowledge is documented in more than 40 pages of business rules and examples of how to handle particular situations. A price list alone is not enough: each request may be subject to different terms and exceptions.
A language model can miss a condition or apply it inconsistently when handling many rules at once. That is why we built a hybrid workflow: AI interprets the request; traditional automation applies the rules and calculates the quote.
Those rules are written in code, with each condition and calculation defined explicitly. When the rules are implemented and tested properly, the same inputs and conditions produce the same result. We can trace how a price was calculated and stop the process if information is missing. We do not let AI make up the terms of a sale.
There are four steps:
- Interpret and complete. Identify what is needed and, on WhatsApp, ask for missing details.
- Check. Verify products, prices and rules. Stop if information is missing; refer requests that need technical assessment to a specialist.
- Prepare. Apply the relevant conditions and create a draft in the business system.
- Review and send. A salesperson checks the quote, makes any corrections and approves it for sending.
Fig. 1From request to quote
- 01 Interpret
Reads the request and structures data.
- 02 Check
Checks data, products and prices.
- 03 Calculate
Applies rules and drafts the quote.
- 04 Approve
Checks, corrects and approves.
02
Business rules are part of the solution
Before automating, you need to explain how the team makes decisions in specific situations. For example, if the system recommends an accessory the customer did not request, it should list it as an optional extra, without adding its cost to the price of what the customer asked for.
The same approach works for a distributor: AI identifies a product even when the customer uses a different name for it, while the catalogue and commercial terms determine the price. In logistics, AI interprets a request for transport and the system calculates the rate based on the destination, weight and type of service.
The common pattern is recurring demand, information arriving from third parties in different forms, and established internal rules. Reducing the work per request allows the team to handle more requests without costs rising at the same rate.
03
Using the time saved to handle more requests
Today, the system produces drafts for around 30% of quote requests. To cover more requests, the system needs rules that capture more of what the sales team knows.
As we explained in From individual productivity to EBITDA impact, saving time is only the first step: you then need to decide where the team can put that time to best use.
In cases like this, spending much of the time saved on improving the system helps it handle a wider range of requests. The sales team reviews exceptions and clarifies the rules so that more quotes can be prepared automatically.
In our experience, critical processes like this call for three to six months of work with the team to define the rules in detail, test them against real cases and refine how exceptions are handled. The more up-to-date documentation the team has at the start, the less time it needs to spend explaining how things work. That work should be planned from the outset.
Every rule added and validated can be used for future quotes. Once that knowledge is built into the system, the team can handle more requests with the same resources.
In a follow-up edition, we will return to the commercial results: quotes sent, contracts won and sales.
If you recognise a similar pattern in your business, start by asking your team to write down what they know about the process: the business rules, the conditions they apply, the exceptions and real examples. That lays the groundwork for effective automation.
