AI SaaS Founders: Hire FDEs Before CSMs or Watch Deployment Kill You
# AI SaaS Founders: Hire FDEs Before CSMs or Watch Deployment Kill You Jason Lemkin, founder of SaaStr and former EchoSign CEO, has a framework for the FDE vs CSM hiring debate that cuts through the usual CS playbook. The question: should AI agent companies hire more Forward Deployed Engineers or Customer Success Managers? Lemkin's answer: it depends on where your bottleneck actually is. In AI B2B, deployment is the new constraint, not retention. ## The Real Diagnostic Hire FDEs first if: - Deals close but time-to-value is 60+ days - Customers train agents themselves and hit 40-50% of potential performance - You're seeing silent churn: customers sign, go quiet, disappear - NPS gets dragged down by "couldn't get it working" feedback Hire CSMs if: - Agents are already live and performing for most customers - Churn happens at renewal despite successful deployments - You have predictable, repeatable onboarding that doesn't require customisation - Most customers hit their goals in the first 30-60 days ## The Sequencing Play Lemkin's actual recommendation: don't choose. Sequence it. Get one strong FDE embedded with your top 3-5 customers. Document what they do. Systematise the training. Then hire CSMs to maintain those relationships at scale. The worst outcome: massively scaling CS before deployment is solved. You are just hiring people to manage unhappy customers. ## Why This Matters for ANZ Sales Teams This is the FDE vs CSM question reframed for AI products. Traditional SaaS let you scale CSMs early because onboarding was predictable. AI agent products have a deployment problem that looks like a retention problem. For sales teams selling AI tools: if your close rate is strong but your customers aren't going live, this is your signal. The quota stays the same whether customers deploy or not, but your comp next quarter depends on renewals. Lemkin built EchoSign to $100m ARR and sold to Adobe. He runs a $90m venture fund and the largest B2B/SaaS founder community globally through SaaStr. His take on this comes from seeing hundreds of companies make this exact mistake: scaling sales, then wondering why the metrics break. The framework is simple. The execution is not. But if you are selling AI SDR tools or any AI agent product, deployment velocity determines whether your patch stays viable.