Forward Deployed Engineering Principles at Ramp

The Discipline of Scoping

Forward Deployed Engineering (FDE) at Ramp prioritizes deep context gathering over reflexive agreement. When faced with urgent enterprise requests, such as a Friday-night demand for an SAP S4 HANA integration, the team pauses to validate the necessity of the request. Engineers must distinguish between genuine customer requirements and artificial urgency driven by sales quotas. Failing to scope properly leads to wasted engineering cycles, such as the time the team spent building cross-platform mobile features only to discover the client mandated iOS-only usage for all employees.

Scaling Intake with Agents

To maintain velocity, the team automates the request intake pipeline using agents built within Notion. The process transforms high-variance, low-quality input from account managers into structured specifications. The agent performs multiple rounds of back-and-forth questioning with the requester to clarify requirements before creating a spec. This automation reduced the latency of replies from hours or days to seconds and reclaimed approximately 20% of the time previously spent on manual scoping.

Building the Agent Factory

Success in the next 6 to 12 months depends on building an agent harness that manages the entire lifecycle from request to implementation. This requires:

• Implementing rigorous evals using rubrics and human feedback to ensure output quality.
• Grounding agents in internal knowledge bases, including product documentation, help articles, and historical request data.
• Maintaining human judgment to oversee the final output, preventing the creation of a token-maxing slop cannon.

Combining manual scoping with automated agent factories is essential, as failing to automate allows agent-native competitors to outpace the team, while failing to scope leads to building the wrong product efficiently.