Ramp's FDE playbook: scope harder, then automate everything you just scoped.

Ramp's forward deployed engineering isn't about blind customer appeasement. The core practice is disciplined scoping: questioning urgency, validating platform assumptions, and evaluating whether a request solves for multiple customers. A painful example: building a mobile feature on two platforms when the customer mandated iOS only. The lesson is that

gathering context before building prevents wasted engineering effort. Simultaneously, Ramp is aggressively automating the FDE workflow itself. They built a Notion agent for their Slack-based request intake that handles scoping questions, reducing reply latency from hours to seconds and saving roughly 20% of scoping time. The vision is a

full agent pipeline handling context gathering, scoping, spec writing, and implementation. The critical insight is that both disciplines must coexist. Scoping without automation creates a manual bottleneck. Automation without scoping creates a "token-maxing slop cannon" — low-quality output at high velocity. The hardest part remains building reliable context

for agents, which is still an unsolved problem. The talk is grounded in real operational experience at a scaling company. Claims about agent impact are supported with concrete examples and measured time savings. The analysis is practical and avoids hype, though the long-term agent factory vision remains aspirational.