I recently worked with a signage manufacturer who, like many, was running their business on an ERP and a collection of messy spreadsheets. These documents were the company’s brain, holding all the Bills of Materials, costing information, and product parameters. The problem is that a spreadsheet is a list, not a system. It can store data, but it cannot enforce the relationships between that data. When you have dozens of substrates, vinyls, inks, and processes, the number of possible combinations becomes impossible for a human to manage without error.

Spreadsheets Create an Illusion of Control

The anti-pattern here is that spreadsheets create an illusion of control. A sales rep looking at a spreadsheet believes they have all the information they need. Yet, the sheet cannot stop them from picking a vinyl that doesn’t adhere to a chosen substrate. It cannot tell them that a certain thickness of acrylic is out of stock or requires a different routing machine. This leads to endless back-and-forth between sales and engineering, fixing quotes after the fact. The time spent determining the true cost for each custom job is pure waste. The final price becomes a guess, not a calculation.

A quote is a promise, not a document.

The first step away from chaos is structure. In a proper CPQ, the product is broken down into a series of decisions. We call these ‘modules’. A module represents a question, like “What substrate material will be used?” The possible answers, acrylic, PVC, aluminum, are the ‘variants’. Instead of starting with an infinite sea of parts, you start with a guided path of decisions. Every item on the final BOM must originate from one of these modules. There are no free-text fields where errors can hide or special notes that require an expert to interpret later.

From Raw Data to Rules with AI Assistance

Building this structure has traditionally been a manual process. This is where AI can dramatically accelerate the work. In our session with the signage company, we uploaded their two main documents: an example configuration and a massive product costing spreadsheet. Then, we instructed an AI agent to start creating modules from this raw information.

The AI read the documents and proposed a structure. In its first attempt, it created separate modules for substrate material and thickness. I rejected this, because the two are dependent. We told the agent to try again, keeping them linked. The second time, it created a single module for substrates and correctly identified the valid thicknesses for each material. The AI doesn't just extract data; it builds a deterministic configurator that makes invalid combinations impossible.

The AI builds the rules; it doesn't get to break them.

This process also uncovered the AI’s ability to use its general knowledge. When it encountered ‘OSB plywood,’ a material not fully detailed in the documents, it correctly inferred its properties. The AI uses the documents as the primary source of truth but can intelligently fill small gaps, all while an expert guides and validates its work. The result is codified knowledge, not just a digitized spreadsheet. It's enough to build a functional quote, even before every official SKU is mapped for ERP integration, that comes next.

The Configurator as a Partner, Not a Gatekeeper

Once the model exists, the configurator enforces compatibility instantly. We built a rule stating that a 'cast' vinyl only works with PVC and OSB plywood. That rule prevents anyone from quoting an invalid pairing. It also automates calculations. A rep enters the sign's height and width, and the system correctly calculates square footage to drive the cost for substrate, ink, and laminate. No more manual math errors.

The system gives the rep a choice. They can use a dynamic form that updates with each selection, or a conversational chat to describe the needs. Both paths lead to the same valid BOM, because both are powered by the same central model. The interface is a preference; correctness is mandatory.

A good CPQ also helps the salesperson be more effective. We can load the AI with ‘skills’ from sales manuals, product guides, or competitive battle cards. If a customer asks about the trade-offs of aluminum versus birch for an outdoor sign, the rep can ask the system directly. The AI gives a concise, accurate answer. It becomes a sales coach embedded in the quoting tool.

If knowledge leaves with people, the business is fragile.

This turns the CPQ from a gatekeeper into a partner. It doesn’t just stop reps from making mistakes; it actively helps them build a better solution and close the deal with confidence.

You make the system more useful than the workaround.

Key takeaways

  • Spreadsheets offer an illusion of control but cannot enforce complex product rules, leading to errors.
  • AI can dramatically accelerate CPQ setup by structuring product data from raw documents like spreadsheets.
  • A configurator should be a sales partner, not a gatekeeper, embedding knowledge to help reps close deals.
  • Codifying knowledge in a system makes a business more resilient and less dependent on individual experts.

If your team spends more time fixing quotes than creating them, it's a sign that your spreadsheets have become a liability. A modern CPQ doesn't just digitize your data; it codifies your expertise into a system that guides reps to the right answer. Think about how a single source of truth for your products could transform your sales process.