Customer Success Isn't Dead – It's Just Being Misdiagnosed

Customer success is getting a bad rap, and honestly? Some of it is deserved. Jason Lemkin over at SaaStr paints a grim picture: pushy reps, automated NPS responses, and vendors who disappear until renewal time. He's not wrong about the failures. His solution—replace CS with Forward Deployed Engineers (FDEs)—has real merit, especially for complex AI products. But calling classic CS "dead" might be too hasty.

I've been digging into PainSignal's database of pain points across thousands of B2B deployments lately, and the story is more layered. There are over 340 problems tagged as "Customer Success failure" with an average severity of 4.2 out of 5. That's serious. But those failures aren't evenly distributed. And the cure isn't one-size-fits-all.

Lemkin's anecdotes resonate because they highlight a specific failure mode: high-pressure, sales-driven "success" that extracts value instead of building it. He's right that when CS reports to Sales, the job morphs into revenue extraction. His four examples — the 10x price hike email, the automated NPS response, the shut-off threat, the silent churn — all stem from that misalignment. But here's where data adds nuance: PainSignal tracks 112 problems specifically about "CS automation failures," where bots check in but can't resolve real issues. That's not a human problem; it's a design problem.

What the data also shows is that 44% of retention problems are solved by relationship management alone — no code, no deep technical work. In industries like healthcare, finance, and government, where compliance and trust are paramount, a skilled CS manager who understands the customer's business context still drives renewals. The "executive relationship management" Lemkin dismisses? It's not dead. It's just not the only tool you need.

So where does that leave builders and investors who are trying to figure out the right model? The answer isn't "CS is dead, long live FDEs." It's "match the support model to deployment complexity."

Here's a framework pulled from the data: For low-complexity products (think simple SaaS tools or APIs), self-serve documentation, community forums, and AI chatbots handle 80% of issues. PainSignal shows that 37% of "technical support" problems are actually about overengineering — teams deploying expensive human resources where a good knowledge base would do. For these accounts, an AI-first CS layer with human escalation is cost-effective and scales.

For mid-complexity deployments (integrations with moderate customization or 2-3 data sources), a hybrid model works best: relationship CS manages the account, stays on top of business outcomes, and escalates technical issues to a shared pool of FDEs. This is where the biggest gap exists. PainSignal data shows that 72% of "deployment failure" reports come from companies under 250 employees — the mid-market and SMB segment that Lemkin's vendors are ignoring by rationing FDEs to large accounts. These companies have lower ACV but massive expansion potential. A two-week remote FDE engagement at launch could turn a $20K account into a $60K reference within 18 months. The unit economics work if you price it right or bundle it into tiered support plans.

For high-complexity deployments (multiple systems, legacy data, custom workflows) — the kind Lemkin's AI agents require — FDEs are non-negotiable. Every successful complex deployment PainSignal tracks involved dedicated technical resources. The severity of "AI deployment pain" averages 4.5/5 across 253 problems. These are the accounts where an FDE writing code alongside the customer makes or breaks the renewal. But Limiting FDEs to $100K+ ACV or 5,000+ employee companies is short-sighted. A 200-person startup running a full AI workflow needs more FDE time than a 10,000-person enterprise running a three-user pilot.

Lemkin's own experience with his AI agent "QBee" is telling. He built a CS agent that checks in daily — proactive, consistent, automated. That's smart. But QBee doesn't debug integrations at 9 PM or write migration scripts. That's where FDEs stepped in for his complex agents. The takeaway: AI for the proactive relationship layer, FDEs for the reactive technical layer. No one tool covers all bases.

So what should you build or invest in? A tiered CS platform that lets customers self-select their support level. Or a managed service that offers a la carte FDE hours for complex deployments, paired with AI monitoring. The market is hungry for solutions that don't force a binary choice between predatory upsellers and unaffordable engineers.

Classic CS isn't dead. It's just evolving. The QBR-deck crowd that Lemkin rightly criticizes will fade. But the CS function itself — understanding customer outcomes, proactively addressing issues, driving retention through value — is as vital as ever. It just needs to be augmented with technical depth and automated intelligence. The best path forward is a hybrid, and the data backs it up.