I had a discovery call last week with someone who used to run a CPQ system. Smart guy. Knows the industry. The kind of person who doesn't ask polite questions to fill silence.
Fifty minutes in, I realized he had asked three questions that, taken together, said more about where this industry is heading than the last six analyst reports I've half-read on flights.
So I'm writing them down. Because if you're in CPQ right now, in any role, on any side of the table, you're about to have to answer all three of them yourself.
He asked this about ten minutes in, watching me do in real time what used to be a 300-hour project.
It's a polite question. The impolite version is: aren't you about to put yourself out of business?
And the answer most consultants give, when asked, is some version of "well, we'll still be needed for the strategic stuff." Which is what every industry says right before it gets restructured by someone who didn't grow up in it.
Here's the actual answer, the one nobody wants to hear at their own annual kickoff. The hourly rate model is going to collapse. Not slowly. Not gracefully. Not for everyone at the same time. But it is going to collapse.
And before anyone reaches for the "but our consultants add irreplaceable value" line, let me introduce you to the translation industry.
In 2005, translation was a 3 billion dollar global market. Computer-aided tools arrived. Neural machine translation followed. A working translator went from 1,000 words a day to 10,000 words a day. Everyone with a translation degree spent the next decade panicking.
The market today is 70 billion dollars.
Twenty times bigger. Not because anyone protected the old margins. The opposite. The price per word fell off a cliff, and demand exploded into it. Suddenly it made sense to translate things nobody used to translate. Internal manuals. Niche e-commerce. Long-tail content for tiny markets that were never worth the effort.
CPQ is sitting on the same slope. The hours per project are about to collapse. The dollars per project will follow. And then the number of projects, the number of buyers, and the size of the addressable market are going to do something that nobody charging 250 euros an hour is currently planning for.
The revenue model question is easy to ask. It's the hardest one to sit with. Because the honest answer is that some firms are about to shrink, some are going to disappear, and some are going to be ten times bigger in three years.
Which one you become depends entirely on whether you're protecting the old margin or chasing the new market. You can't do both. Pick.
This one sounds technical. It isn't. It's a vibes check.
What he was really asking: is the AI part real, or have you just wrapped a chatbot around the same old constraint solver and slapped a new logo on it?
Fair question. There's a lot of that going around.
Here's the truthful answer for any serious CPQ tool, including ones being marketed as if they invented configuration last Tuesday: yes, there is still a rules engine. There has to be. If there isn't, run.
A pure language model configurator is going to hallucinate. Not maybe. Not occasionally. It is going to confidently quote a chassis with the wrong engine, a hydraulic system that physically cannot route through the cabin, or a delivery date the factory will laugh at. You cannot ship that to a customer. You cannot put your name on it. You cannot survive doing it twice.
But the inverse is just as bad. A pure rules engine, the kind we've been lovingly hand-crafting in this industry for 25 years, is going to keep doing what it's always done. Ask the customer 47 dropdown questions when they wanted to upload a PDF, describe their problem in their own words, and get a price by tomorrow morning.
The shift is not AI replacing rules. Anyone selling that is either confused or lying.
The shift is AI on top of rules. The language model handles the conversation, the recommendation, the messy translation from "I need something that can run at minus 40 degrees on diesel and survive a Norwegian winter" into a structured configuration. The rules engine does what rules engines have always done, which is guarantee that whatever comes out the other side is actually buildable.
This is not a compromise. It is the architecture. And the firms that try to skip one half of it, in either direction, are going to spend the next two years finding out why both halves exist.
This was the most revealing question of the three. And the most polite. Because what he was actually asking was: am I the buyer for this new wave, or am I too small to matter to you people again?
The honest historical answer is: yes, you were too small. CPQ was sold to companies above a certain revenue line, with a certain product complexity, willing to fund a year-long project for a return that wouldn't show up until two years after the project ended. The sales motion required deeply technical reps flown in from another country. The implementation required senior consultants charging consultant rates. The buyer needed enough pain to justify all of it before signing.
Most companies took one look at the spreadsheet and quietly went back to Excel.
That math is breaking. Hard.
When a working configurator can be stood up in a week, from a product brochure, by the company's own subject matter expert using a chat interface, every variable in the old equation changes. Sales motion. Buyer profile. Deal size. Time to value. The complexity threshold at which CPQ becomes worth doing at all.
The companies that were "too simple" for traditional CPQ are exactly the companies that become addressable when the cost curve breaks. Not because they suddenly became complex. Because they can finally afford the version of CPQ they always wanted. A guided online configurator. A clean quote. Tracking on whether the customer actually opened the proposal. The ability to maintain the whole thing themselves without booking a consultant for the privilege of changing a price.
Markets don't grow by selling more of the same thing to the same buyers at the same price. They grow when the price collapses far enough that an entire layer of buyers, the ones who were always there, the ones who always wanted in, can finally afford the door.
Read them again. Side by side.
How is your revenue model going to change? Is the AI part real, or are you faking it? Am I the buyer now, or am I still too small?
They are the same question, asked from three angles. Has the economics of this industry actually changed, or is this just better marketing on top of the same old projects?
The economics have changed. The translation market grew twenty times when its cost curve broke, and CPQ is on the same curve now, whether the incumbents are ready or not. The technical winners will be the firms that combine language models with constraint solvers instead of picking a side. The commercial winners will be the firms that notice the buyer profile is shifting down-market and stop pretending the only real customer is a 500 million dollar manufacturer with an eight-figure IT budget.
The next two years are going to sort the firms that understood the shift from the firms that priced their services as if nothing happened. It will not be a gentle sort.
If you're in CPQ right now, on any side of the table, the three questions above are coming for you. The only thing you get to choose is whether you have an answer ready, or whether you find out what your answer is in real time, on a call, in front of a customer who already knows what yours should be.
I know which side I'd rather be on.