The Real Customer Success Crisis Isn't What SaaStr Told You

There are two parallel universes in B2B right now. In one, customer success is being rebuilt from scratch by the fastest-growing AI companies—Lovable, Harvey, Assembly AI—and the old playbook is "a liability." In the other, customer success teams are still fighting a war against green-screen terminals, HIPAA audits, and ERP systems that predate the internet. And that second universe is way bigger than most people realize.

Jason Lemkin's recent piece on SaaStr captures the first universe brilliantly. The panel at SaaStr AI 2026 was an autopsy and a rebirth—Monica Perez replacing Gainsight with a custom command center on Lovable, Tom Ronen running old-school change management at an $11B legal AI company, and everyone agreeing that NPS is dead. The data from those companies is impressive: Lovable hit $400M faster than any company before it, Harvey crossed $190M ARR, and forward deployed engineering roles are exploding.

But here's what the SaaStr conversation missed. The advice from that stage—rename CSMs to forward deployed engineers, rebuild your stack weekly, stop saying "AI-powered"—assumes you're a venture-backed AI startup with clean modern infrastructure. Most of the B2B world isn't that.

Our data tracks 19,860 customer success problems across 84 industries. The average severity is 3.7 out of 5—meaning these aren't minor annoyances, they're structural barriers. And the most painful problem, far and away, isn't activity metrics or NPS. It's legacy system integration.

We've logged 1,876 problems specifically about integrating customer success tools with on-prem ERP and CRM systems. That's nearly 10% of all CS problems we track. The average severity: 4.6 out of 5—higher than any specific CS activity problem the SaaStr panel discussed. For companies in healthcare, manufacturing, and logistics, the AI-native future isn't just years away; it's structurally blocked by systems that can't even expose an API.

The "CSM title is a tax" argument from Assembly AI's Ryan Seams makes perfect sense at a company selling to technical buyers who respond to "forward deployed engineer." But try renaming your customer success roles at a medical device company that requires HIPAA-certified personnel. Or at a financial services firm where your account team needs FINRA licenses. Our compliance data captures 743 problems tagged to compliance requirements in customer-facing roles, with an average severity of 4.3. That certification overhead doesn't disappear because you change a title.

The SaaStr panel correctly noted that "over half of CSM activity has zero correlation with retention." But our broader dataset shows that technical account manager problems (2,899 logged, severity 3.4) still outnumber forward deployed engineer problems (1,209 logged, severity 3.8) by more than 2 to 1. The shift the panel described is real at the leading edge, but it hasn't penetrated most industries yet. The FDE role is still a novelty in pain reporting—possibly because it hasn't hit scale in the compliance-heavy sectors that employ most CS professionals.

What does this mean for builders and investors? The AI-native CS transformation is a real opportunity, but it's not a one-size-fits-all revolution. There are two distinct markets emerging.

Market one: AI-native companies with greenfield infrastructure. These are the Lovables and Harveys where teams can rebuild their entire CS stack weekly, track every customer with individualized bots, and align adoption directly to consumption-based revenue. This market is growing fast and it's where most of the SaaStr advice applies. Investment thesis: tools that enable this kind of custom-built, versioned CS platform.

Market two: legacy-regulated industries with complex compliance burdens and decades-old systems. This is healthcare, manufacturing, banking, legal (the old kind), logistics. Here, the CS playbook can't be rewritten from scratch; it has to be carefully adapted around integration constraints. Investment thesis: middleware that bridges modern CS tools with legacy systems, or compliance-friendly CS platforms that don't require a full rebuild.

The SaaStr piece is right that NPS is dead—our data shows NPS-specific problems have severity 2.1, far below the platform average. And the panel's point about adoption being the starting line, not the finish, is spot-on. But the framing that customer success is being "rebuilt" assumes you had something modern to rebuild from. Most companies are still trying to get their first CS platform to talk to their order management system.

Lovable's Monica Perez said the team closest to the customer should build the tool. That's a great philosophy when your tool is Lovable and you're not technical. But for a manufacturing company with 15-year-old SAP installations, building your own CS platform isn't a weekly side project—it's a multi-quarter initiative with compliance reviews.

The most valuable takeaway from the SaaStr piece isn't the specific playbook; it's the mindset. AI should be ambient, not the headline. The clock needs to match product speed. Value and revenue should be the same thing. Those principles apply everywhere, even if the implementation looks different in a hospital system than at a dev tool startup.

But as a builder or investor, know which world you're serving. The AI-native CS playbook is powerful but narrow. The broader opportunity—helping the other 80% of B2B inch toward that future—is where the real pain lives. And based on our data, that pain is deep, widespread, and nowhere near solved.