AI Pricing Infrastructure Is a Bigger Bottleneck Than You Think

I just read Jason Lemkin's latest piece over at SaaStr, where he makes Nue his AI App of the Week. If you haven't seen it, he argues that the hardest problem in B2B AI right now isn't building the product—it's pricing it. And he's not wrong. The whole per-seat model is crumbling as AI agents do the work humans used to do, and companies are scrambling toward consumption, credits, and outcome-based models. But the tooling to actually quote, bill, and recognize revenue on those new models? Mostly a mess of duct-taped legacy systems.

Nue's pitch is to collapse that stack into one platform where pricing changes become config changes, not engineering projects. It's a compelling argument, and Lemkin makes it well. But I think he missed a layer of the story that our data makes painfully clear: this isn't just an enterprise headache. It's hitting smaller, agile AI companies hardest—and the severity numbers are eye-opening.

PainSignal tracks problems reported by builders, and in this space, we're seeing 47 pricing-related problems with an average severity of 3.8 out of 5. That's not theoretical. That's real friction, right now. And when you zoom in on usage-based pricing specifically, it gets worse: 128 problems, with a 32% quarterly growth rate. The trend Lemkin describes is real, but the pain is accelerating faster than most people realize.

What he doesn't address is the SMB and mid-market side of this. Nue is an enterprise-grade platform, built on the Salesforce data model. That's great if you're a well-funded scale-up with a RevOps team. But we've tracked 28 problems specifically from SMB B2B AI companies struggling with pricing infrastructure adoption. Their average severity? 4.1 out of 5. These are teams that know they need something better but are getting squeezed by cost and complexity. The market is screaming for a lighter-weight solution, and I don't think Lemkin's piece acknowledges that gap.

Then there's the API-first AI startup crowd. These companies are building on top of LLMs and reselling that compute, which means they're billing customers for GPU or API usage—often with razor-thin margins. Our data shows 19 problems in this subcategory alone, tagged under AI API billing and cost passthrough, and the average severity is a brutal 4.4 out of 5. That's the highest we track in this entire category. If Nue or any other platform can't handle real-time metering with enough granularity to squeeze out margin, these startups will keep building their own—and likely doing it badly.

Lemkin's piece does reinforce the big-picture trend, and our data backs him up. But I want to push back on one claim: that every AI company is walking away from pure per-seat. Not so fast. Our data shows 23% of AI SaaS problems still involve per-seat or per-user pricing as a primary component, often hybridized. Enterprise tools with human-in-the-loop workflows aren't ditching per-seat entirely; they're just layering on consumption elements. Pure per-seat is declining, but it's not dead, and any pricing infrastructure play needs to handle that hybrid reality without treating it as an afterthought.

For seed investors, this is where the signal gets interesting. If you're backing AI-first startups, ask them how they're handling pricing today. If they say "Stripe and spreadsheets," know that's a time bomb. And if they're API-first, ask how they're doing metering and passthrough. The ones with a clean answer might have a moat in the making, because most of their peers are drowning here. Infrastructure that solves this isn't just a back-office tool; it's becoming a competitive advantage. Lemkin's right about that.

The PainSignal data also reveals something else: latent demand for exactly what Nue is doing. We've seen a surge in related app ideas, like a unified usage-based billing and revenue recognition platform. Builders are coming up with this problem on their own. That means the market is ripe for a well-executed solution, but it also means the window is open for competitors—especially ones that nail the SMB and API niche.

Bottom line: read Lemkin's article for a sharp take on the problem and a solid founder perspective. But if you're an investor or a builder trying to figure out where the real pain lives, the story is deeper than that. The severity and growth rate of pricing infrastructure problems tell me we're in the early innings of a shift that will create real winners—and leave a lot of promising startups stuck in spreadsheet hell.

Check out Jason Lemkin's full article here for his take on Nue, and then come dig into the 47 pricing problems builders are reporting —plus the 128 usage-based pricing headaches growing at 32% a quarter.