CPQ for manufacturing is the configure, price, quote process that turns a customer request for a configurable product into a priced, buildable quote. We think most manufacturers do not need a bigger rules engine. They need something that can read the request in the first place, because that is where the days actually go.
A request for a configurable assembly rarely arrives as a configuration. It arrives as a drawing, a customer specification that points at three other specifications, a spreadsheet of line items with no article numbers, or an email that says "same as last year, but 3 metres longer and in stainless". Before any pricing rule can fire, a person has to read all of that and work out what was asked for.
At Lleverage we build agents that do that first reading and hand a configured draft to the person who signs it off. Our order management work covers quote generation alongside order intake, because in most back offices the two arrive through the same mailbox and get handled by the same three people. To see it run against your own request files, book a demo.
What is CPQ for manufacturing?
CPQ for manufacturing is the set of rules and workflows that take a customer request for a configurable product, validate the configuration against what the factory can actually build, price it, and produce the quote document. It exists because configurable products have too many valid combinations to hold in a price list.
The classic case is engineered-to-order equipment: a machine with 14 option groups, some of which exclude each other, priced off a bill of materials that changes with every choice. No salesperson holds that in their head. A spreadsheet version of it drifts within a quarter. CPQ moves those constraints into a system, so an invalid machine cannot be quoted and a valid one is priced the same way every time.
Two things separate manufacturing CPQ from the generic version sold to resellers. The first is that a configuration has to be manufacturable, not merely sellable, so the rules reach back into engineering data, lead times and material availability. The second is the output. It is rarely one number: a quote document, a preliminary bill of materials, sometimes a drawing, and a delivery date production has to be able to honour.
Why does quoting a configurable product take days?
Quoting takes days because most of the elapsed time is not calculation, it is interpretation. Someone has to read an inbound request in whatever shape it arrived, translate it into the company's own part numbers and option codes, chase the two things the customer left out, and only then start configuring. The rules engine runs in seconds. The reading runs in days.
Three things stretch it further. Requests queue behind the one or two people who know the product range well enough to interpret them, so quoting speed is bounded by a diary rather than by capacity. Clarification questions add a round trip each, so the desk learns to avoid asking them. And none of the interpretation is written down, so the same translation gets done again the next time that customer sends the same badly formatted request.
That last one is the quiet cost. Product knowledge sits in the heads of the people at the desk. A colleague cannot pick up the request, and a new hire takes months to become useful. At SIG Benelux, a building materials distributor running Dynamics 365, roughly 3% of inbound order lines carried an article number and about 47% arrived as description only. That is the same interpretation problem, one step downstream from the quote.
You deliberately leave questions unasked, because asking makes the processing take longer. You would want to ask, but there is no time. That is being made possible now.
Patrick Wijers, Sales Desk Manager at SIG Benelux, describing the order desk before the work started. Quoting desks say the same thing about clarification emails.
What do configure, price and quote actually involve?
Configure validates the combination of options against engineering and production constraints. Price applies cost build-up, discount policy and margin rules to the validated configuration. Quote produces the document the customer receives, with the commercial terms, lead time and any drawings or specifications attached. Each stage has a different owner and a different failure mode.
| Stage | What it decides | Where it goes wrong |
|---|---|---|
| Configure | Which option combinations are buildable, and with what components | Rules drift from what the factory can actually make; engineering fixes them by email |
| Price | Cost roll-up, discount authority, margin floor | Sales prices from an old sheet; margin is only checked after the order lands |
| Quote | The document, lead time and validity period | Lead times are copied from the last quote rather than from current capacity |
| Interpret | What the customer actually asked for | Not owned by any system, so it falls to whoever opens the email |
The fourth row is the one most CPQ projects leave out. Configure, price and quote all assume a structured request already exists. In an SME manufacturer it usually does not, which is why a well-implemented rules engine can still sit behind a five-day quoting cycle.
Where does traditional CPQ software stop?
Traditional CPQ stops at the point where input has to be understood rather than validated. It is built to check a configuration a salesperson has already entered, so it needs the options selected, the quantities typed and the customer matched before it can do anything. Everything upstream of that stays manual.
That boundary is reasonable for a business selling through a configurator on its own site, where the customer does the entering. It holds up badly for an SME manufacturer. There, enquiries arrive as PDFs, drawings and forwarded email chains from procurement departments with their own formats and no intention of changing them.
The second limit is maintenance. Configuration rules encode what the factory can build, and the factory changes: a supplier is swapped, a material goes on allocation, a variant is retired. Keeping the rule set true is real work. When it lapses, the system quietly starts quoting things production then has to argue about. In our experience the rule set is accurate at go-live and roughly 18 months behind two years later, because nobody owns it once the project team leaves.
How do AI agents change manufacturing quoting?
AI agents change quoting by taking on the interpretation step that CPQ assumes is already done. An agent reads the request in whatever shape it arrives, matches the described items against product and customer data, drafts the configuration, and hands it to a person to check. The rules engine still decides what is buildable, and a person still signs the quote.
The mechanism is ordinary once you see it. Years of quotes and orders already contain the translations the desk does by hand, so the mapping from a customer's wording to your article numbers can be learned from history rather than maintained as a rule table. Lleverage's order management page puts complex quote drafting at 75% faster when configuration, pricing and document generation run as one flow, with the review kept in place.
Two published examples show the shape. At SPL Treatments, an aerospace finisher in Sheffield, deciding how to treat a first-time part meant reading 60-page customer specifications that point at other specifications, then cross-checking a library of more than 800 standards. It took 30 to 60 minutes. It also took Grant Riley, because nobody else could do it. With the library indexed and an agent walking the references, it now takes 2 to 2.5 minutes, and every recommendation cites the section it came from. That returns roughly 5 hours of senior engineering time a day.
At SIG Benelux the same reading step was automated one stage further down the funnel. 73% of incoming order emails now become a draft order in Dynamics 365 on their own, and 88% of order lines match the correct article on the first try, including lines that arrive with a description and no number. The desk reviews instead of retypes, which is what freed up time for quotes in the first place.
Neither of those is a configurator. Both are the interpretation layer that a configurator needs and does not have. Our view is that this is where the quoting gain now sits for most SME manufacturers, because the rules engine problem was solved 20 years ago and the reading problem was not.
What should manufacturers look for in a CPQ system?
Look for three things: whether it can accept an unstructured request, whether it writes to the system you already run, and who maintains the rules after go-live. Those three decide whether the quoting cycle actually shortens, because a configurator that only accepts clean input moves the bottleneck rather than removing it.
| Requirement | Why it decides the outcome | Question to ask the vendor |
|---|---|---|
| Reads unstructured requests | Most enquiries arrive as PDF, drawing, spreadsheet or email prose | What happens to a request with no article numbers? |
| Writes into the ERP | A quote that has to be retyped into the ERP loses the time it saved | Which ERP nodes do you write, natively or through a connector? |
| Keeps a person in the loop | Wrong quotes cost margin and credibility, and silence is worse than a flag | What does the system do when it is not sure? |
| Learns from corrections | Rule tables decay; corrected examples compound | Does a correction change the next result, or only this one? |
| Shows its reasoning | Engineering has to be able to check a configuration it did not make | Can a line be traced to the data it came from? |
Ask about the unhappy path specifically. A demonstration on a clean configuration tells you very little, because that case was already fast. Bring a real enquiry from your own mailbox, preferably an awkward one. Watch what happens to the parts it cannot resolve. Handing back a flagged gap is a good answer. Filling it in confidently is not.
On integration, check whether the connection to your ERP is native or a file drop, and what happens when the ERP has no usable interface at all. SPL's system of record is a legacy desktop application with no API, and the workflow was designed around that constraint rather than blocked by it. Our integrations cover SAP, Business Central, Exact, AFAS and Infor, and the honest answer for anything older is that the output has to be something a person can carry across.
How long does it take to get quoting automation live?
A narrow quoting workflow reaches production in weeks rather than quarters, provided the scope stays narrow. The SIG Benelux project ran as a 4-week engagement covering one location, one process and one mailbox, and went live on real orders in June 2026. Widening the scope is what turns a 4-week project into an 18-month one.
Pick the request type that is both high volume and painful, not the one that is most interesting. Usually that is repeat configurations from named accounts, where the history is deep and the variation is bounded. Prove the reading step there, let the corrections accumulate, then extend to the harder enquiries once the model of your product range has something to stand on.
Expect the first weeks to be about your own data rather than the technology. The recurring surprise is that quote and order history turns out to be a better source of truth than anyone expected, while master data gaps are the real constraint. If your item data is thin, master data management comes first. It pays for itself across every process that touches the item file.
Frequently Asked Questions
What does CPQ stand for in manufacturing?
CPQ stands for configure, price, quote. In manufacturing it covers validating that a requested product combination can actually be built, pricing it from the resulting bill of materials and discount rules, and producing the quote document with lead time and terms. The scope usually reaches into engineering data rather than stopping at a price list.
Is CPQ the same as a product configurator?
No. A product configurator handles the configure step: which options are valid together and what components result. CPQ wraps pricing, approval and document generation around it. Many manufacturers own a configurator inside their ERP or PLM and call the surrounding spreadsheet work CPQ, which is where most of the quoting delay hides.
Can AI replace a CPQ rules engine?
We do not think it should. Configuration rules encode what your factory can build, and that is a deterministic question with a right answer. What AI adds is the step before: reading an inbound request, matching it to your data and drafting the configuration, so the rules engine gets clean input instead of waiting for a person to type it.
How much does CPQ for manufacturing cost?
Traditional CPQ is normally licensed per sales user with a separate implementation cost, so the total tracks the size of the sales team and the complexity of the rule set. Our own agents are priced per agent rather than per seat, with integration, AI usage and ongoing improvement inside the monthly price. The pricing page explains how that is scoped.
What is the first process to automate in quoting?
Start with the interpretation of repeat enquiries from your largest accounts. The volume is predictable, the history is deep enough for matching to work, and the failure cost is low because the desk still reviews every draft. Once that is running, extend to first-time configurations, where the reading is harder and the time saved per request is larger.
Where the quoting days actually go
The quoting cycle in most SME manufacturers is slow for a reason the business case never shows. Every request has to be read and translated by one of a handful of people before any system can help. That queue is invisible in every CPQ evaluation we have seen. This is where we land: fix the reading, and the configurator you already own starts to look a lot faster.
If you want to test that on your own enquiries, bring 10 recent requests, including the badly formatted ones, and book a demo. We will run them and show you what comes back, including the parts the agent flags rather than answers.