AI CPQ and the Incumbent's Dilemma: What Defensive Posts Tell You

There is a moment in every technology shift that I have learned to watch for. It is not a product launch or a funding round. It is quieter than that. It is the moment when the incumbents stop ignoring the new thing and instead start explaining, at length and in public, why it cannot possibly work.

I have collected those posts for years. From the biggest platforms in enterprise software down to the smallest niche vendors. I have written a few myself, which is a confession we will get back to. And lately, as AI CPQ has gone from curiosity to category, my feed is full of them again. Long thoughtful posts. Confident diagrams. Detailed explanations of why the new tools will never manage what the old ones do.

Every time one scrolls past I think about the German car manufacturers.

The most expensive eye-roll in industrial history

For the better part of a decade the German car industry had a ready answer for electric cars. The range is a joke. The charging network does not exist. Customers do not want them. Batteries are an environmental scandal waiting to happen. And the classic, delivered with a knowing smile at every trade fair: an electric car is just a skateboard with a laptop on it.

Real cars are hard. We have a hundred years of engineering in the drivetrain. That is the moat.

Here is the uncomfortable part. Every single one of those statements was true when it was said. The early electric cars really were worse on almost every axis the incumbents measured. And the incumbents really did have a hundred years of drivetrain engineering.

It just turned out the market stopped buying drivetrains.

Volkswagen was the best-selling car brand in China for roughly four decades. Then the native electric brands arrived - companies that never built a combustion engine and never wanted to. They did not compete on the incumbents' axes. They made the car a software product: the screen, the assistant, the updates, the buying experience. By 2023 BYD had taken the crown in China and the German brands were suddenly fighting for relevance in what used to be their most profitable market.

Not because their engineering got worse. Because the definition of a good car changed under their feet while their best people wrote very convincing essays about charging infrastructure.

And here is the detail I find most telling. The incumbents did build electric cars in the end. They bolted batteries onto platforms designed for engines, shipped software the way they wrote software for door controllers and wondered why the natives kept winning. Being right about how hard the old thing is turned out to be something completely different from being good at the new thing.

The AI CPQ shift is running the same movie

Our little CPQ industry is compressing those twenty years into about two.

The AI-native CPQ platforms have arrived. They are designed for language models the way native electric cars are designed for batteries: from the ground up, data structures and all.

And right on schedule the defensive posts have started. "The demos are toys". "Production is harder than it looks". "We have spent years building what these startups cannot imagine".

Just like the German engineers, the posts are largely correct. A prototype is not a product. Production is genuinely hard. I have written a whole article about why a slick AI configurator demo is not the same thing as a production system and I stand by every word of it.

But correctness is not the point of those posts. The point is the timing. Nobody writes a thousand words explaining why a competitor cannot work while that competitor is not working. You write it when your pipeline starts asking questions. The defensive post is not analysis. It is a pain signal, published voluntarily.

There is usually a second tell and it is my favorite. The same voices arguing that the new approach cannot survive the real world will in the next breath announce their own AI initiative. On the roadmap. In early access. Coming with the next release. So the new thing is impossible and on its way, at the same time. The German brands did exactly this - mocking electric cars in one press release and announcing their electric strategy in the next. When a vendor does both in the same quarter you are not reading a technology assessment. You are watching someone hedge in public.

What AI CPQ actually means (and what it does not)

Since we are here, let us be precise, because the term is already being stretched by marketing departments.

AI CPQ does not mean a chatbot in the corner of a legacy configurator. That is the batteries-in-an-engine-bay version. Real AI-native CPQ means the platform was designed for LLM reasoning from day one. The product data is structured for a language model to reason over. The interface is a conversation as much as a form. And critically, the intelligence is paired with a guarantee: a symbolic constraint solver that makes invalid quotes impossible, no matter how creative the language model gets. I have written more about that architecture in "Can You Trust an AI-Generated Quote?" but the short version is simple. The LLM is the car, the solver is the guardrails. You need both.

That combination changes what CPQ is measured on. Which brings us to why the natives keep winning.

Why the natives win anyway

Here is what the incumbents get wrong, in cars and in CPQ. And it is subtle enough that smart people miss it for a decade.

They believe the moat is the accumulated difficulty of the old thing. All those years of platform hardening, integrations and edge cases. That accumulation is real. But a technology shift does not attack your moat. It changes what lake the customers are swimming in.

Look at the new axes buyers actually measure AI CPQ on. Time to live: weeks, not quarters. Price: something that makes sense for a fifteen-million-euro manufacturer, not only for the enterprise tier. Interaction: describe what you need in plain language instead of learning a form-based wizard. Output: a living proposal with a valid configuration, not a static PDF. The manufacturers below the enterprise tier, the ones I wrote about in my piece on the CPQ glass floor, do not care that a legacy platform carried the biggest enterprise rollouts of the 2010s. That was the old axis. Being unbeatable on the old axis is how you lose slowly while feeling like you are winning.

And the bolt-on trap is just as real in software as it was in Wolfsburg. You cannot retrofit a legacy configuration engine for LLM reasoning any more than you can turn a combustion platform into a good electric car. The data structure is the platform. The natives designed for it. The incumbents are shipping batteries in an engine bay and calling it transformation.

What I tell people who ask

I promised a confession, so here it is. I ran enterprise CPQ implementations for twentyfive years. My own business model is being disrupted by the same wave I am describing. And the 2019 version of me would have written those defensive posts too. They would have been well argued, full of true statements and completely beside the point. I know, because I remember what I told customers about early configurator startups back then.

If you are a buyer: read the defensive posts carefully, because the technical objections in them are often valid. Then notice what the existence of the post tells you about where the market is going. Pressure-test the natives hard on the unglamorous problems. But do not let an incumbent's essay about production readiness convince you the shift is not happening. The essay is evidence that it is.

If you are a vendor: I say this with real sympathy, because it is the hardest advice in business. The energy you spend explaining why the natives cannot work is energy the natives spend on working. Volkswagen did not lose China to a bad argument. It lost to a better product on an axis it refused to measure.

The market does not read your posts. It just moves.

AI CPQ: FAQ

What is AI CPQ? AI CPQ is configure price quote software built around artificial intelligence, typically a large language model for conversation and reasoning combined with a constraint solver for validity. Users describe what they need in plain language and receive a valid configuration, price and proposal. The strongest implementations are AI-native, meaning the platform was designed for LLM reasoning from the start.

What is the difference between AI-native CPQ and AI bolt-on CPQ? An AI-native platform is designed for LLM reasoning from day one, including the data structures the AI works with. A bolt-on adds a chat layer to a legacy configuration engine. The difference mirrors native electric car platforms versus batteries retrofitted into combustion designs: the constraint is architectural, not cosmetic.

Are the incumbents wrong that production CPQ is hard? No and that is what makes the pattern dangerous. The defensive arguments are usually technically correct. Production is hard, validity guarantees matter, integrations matter. The mistake is strategic: being right about the old difficulty says nothing about who wins on the new axis customers actually buy on. One way to mitigate the risk is to work with experienced consultants who have seen both sides of the shift.

Why did Volkswagen lose market share in China? Native electric brands like BYD redefined the product around software, batteries and buying experience while incumbents optimized combustion platforms. By 2023 BYD had overtaken VW as China's best-selling brand. The lesson for software markets: incumbents rarely lose on their own axis; the axis moves.

How should a manufacturer evaluate AI CPQ vendors? Test them on the unglamorous problems: invalid combinations refused, pricing logic honored, buildable bills of materials, auditability. Then compare speed and total cost against a traditional project. Skepticism about demos is healthy; skepticism about the shift itself is how markets get lost.

--

Yes, you actually read the entire blog post. Not bad. The reward is that I'll give you a link to test out one of the products redefining CPQ. Here's the link.