Something unexpected is happening at the edges of industrial software. Here’s what we’re seeing from the ground.
I spend most of my time building software companies for manufacturers alongside incredible founders. It is not a glamorous beat. The problems are deeply operational — scheduling, procurement, quality, workforce planning. Not the kind of territory that generates conference keynotes or breathless press releases.
Which is precisely why, when something surprising happens, I pay attention.
Over the last three months, one company in our portfolio has been growing at a pace I haven’t seen in industrial software before. Double-digit week-over-week growth. And not just the numbers — the qualitative signal is just as striking. Users on both sides of the marketplace are expressing a level of satisfaction that is rare in B2B, let alone in manufacturing. Something changed for them. I’ve been trying to understand what.
The company runs a double sided marketplace for factories. I’m keeping them anonymous by choice — they’re building something real and don’t need the distraction — but the pattern is instructive enough to share.
When I sat down with the founding pair to understand what drove the inflection, their answer was immediate: AI changed the fundamental economics of their operation. Not in a vague, aspirational way. In a very concrete one.
The team is small. Deliberately so. They’ve embraced AI tooling fully — writing code, running analyses, building internal apps — and that decision has compounded over time in ways that are difficult to replicate quickly. In the last few months alone, they’ve shipped nearly ten micro-applications that transformed how they run the business. Matching logic. Onboarding flows. Qualification tools. Analytics dashboards that would have required a data team eighteen months ago. Each one took days, not months.
The aggregate effect is a level of operational capability that simply didn’t exist before at this team size and cost structure. And it shows up on the customer side in a way that matters.
Here is the part I find most interesting.
Before AI, running a high-quality marketplace in manufacturing meant making hard choices about whom you served. The top tier of factories — large, well-resourced, operating at scale — were worth the attention because the economics justified the service cost. Everyone else got a diminished experience. The market structure rewarded concentration at the top. In this particular vertical it was very very true — about four weeks of consultant time and data work were the prerequisite.
What this team has done, almost quietly, is break that logic. By compressing the cost of high-quality service through AI-assisted operations, they’re now able to deliver a level of responsiveness, matching precision, and reliability to customers who would previously have been out of reach. Factories that represent the bottom 80% of the market by size are now getting something close to what only the top 5% could access before (at a cost). That is a real structural shift, and it shows up in the growth numbers.
One of the founders has an interesting perspective on all of this. He comes from a Rocket Internet background — he’s built and scaled marketplaces before, which means he understands the mechanics of liquidity, trust, and network density better than most. When we talked through what’s happening in the market, we arrived at the same conclusion from different directions.
In a world where the software layer is rapidly commoditizing — where any reasonably capable team can now assemble functional products at a fraction of the previous cost and time — the durable assets shift . Distribution, trust, proprietary datasets. Not code. Distribution, meaning you have the supply and the demand and the habit of use. Trust, meaning both sides of the market believe the platform will behave reliably and in their interest. These things take time to build. They don’t compress the way software does.
AI, in this framing, is not the moat. AI is the engine that lets you extend your existing moat further and faster. The marketplace player who fully embraces AI can deliver a level of service that legacy competitors, operating on older tooling and older assumptions, simply cannot match right now. That gap will not last forever — it rarely does — but in the window where it exists, it is a real competitive advantage.
There’s a broader pattern here that I want to name, because I think it matters for how we think about where defensible software businesses are being built right now.
We’re watching two ends of the AI spectrum generate durable value, and they look almost opposite. (friendly jab : we’re also seeing quite a few europeans VCs asleep at the wheel and not understanding neither of those).
On one end are companies going deep into data infrastructure — building the pipelines, the labeling systems, the proprietary datasets that make AI outputs better in a specific domain of the company, at scale. These are not glamorous businesses, but the data assets they accumulate are genuinely hard to replicate and the output is a step change from previous systems. The work is often invisible from the outside, which is part of what makes it defensible (call it the difference between the 95% failed genAI projects and the rest, if you will).
On the other end — and this is what feels new to me — verticalized marketplaces are re-emerging as a compelling category. Not because the marketplace model is new, but because AI has changed what a two-person or five-person team can deliver in terms of service quality, operational depth, and customer experience. The combination of network effects and AI-augmented operations creates something genuinely difficult to displace.
Both ends share the same underlying logic: distribution and data are the scarce things. Software is not.
One company growing fast does not a trend make, and marketplaces have famously failed in manufacturing before, often because the trust problem was underestimated, the liquidity challenge was brutal and the service equation too hard. The structural difficulties haven’t disappeared.
But I think the conditions have changed enough that the bet deserves a second look. The cost of building and operating a high-quality marketplace has dropped substantially. The gap between what an AI-enabled team can deliver and what a legacy player delivers has widened, at least temporarily. And the users — manufacturers, suppliers, operators — are increasingly ready to transact digitally in ways they weren’t three years ago.
When a founder with genuine marketplace experience tells me that AI changed the fundamental unit economics of his business, I take that seriously. Not as a prediction, but as a signal worth following.
We’re paying close attention.
Time to build.