# How Forward Deployed Engineering is done at Decagon — Sunny Rekhi

Source: [AI Engineer](https://www.youtube.com/watch?v=7wu2hsRfvV0)  
Feed7 permalink: https://feed7.dev/p/how-forward-deployed-engineering-is-done-at-decagon-sunny-rekhi-02myrh1  
Published: 2026-07-28T17:00:35.000Z  
Trust: Source Linked (source_linked)

## Why Included

Decagon splits deployment between configuring each customer’s agent and turning repeated enterprise requests into product features. The scarce skill is resisting brittle one-offs.

## Source Summary

Decagon describes two deployed functions: configure the customer’s agent and turn field requests into reusable product work. After roughly **25 custom integrations**, it built a self-serve path; the company also grew from **50 to 500 people** in a year.

## Practical Implication

Builders should record the desired outcome before implementation, prove value on a narrow workflow, and inspect every custom request for reuse. If engineers repeatedly configure the same behavior, move that capability into natural-language or self-serve product controls.

## Agent-Ready Context

Decagon describes two deployed functions: configure the customer’s agent and turn field requests into reusable product work. After roughly **25 custom integrations**, it built a self-serve path; the company also grew from **50 to 500 people** in a year.

Builders should record the desired outcome before implementation, prove value on a narrow workflow, and inspect every custom request for reuse. If engineers repeatedly configure the same behavior, move that capability into natural-language or self-serve product controls.

The approach depends on distinguishing repeatable needs from genuinely customer-specific ones. Productizing too early can overgeneralize a single account’s constraints, while moving fast with patches can leave customers owning a brittle black box.

## Connected Context

Feed7 judgment across 262 accumulated Signals:

This makes repeated customer configuration a concrete product signal: prove a narrow outcome, detect recurring requests, then replace bespoke engineering with reusable controls. It confirms forward deployment as a route to product leverage while defining its central judgment call—when evidence is sufficient to generalize—because premature productization and prolonged patching create different forms of brittleness.

- [How Forward Deployed Engineering is done at Kepler — Vinoo Ganesh](https://feed7.dev/p/how-forward-deployed-engineering-is-done-at-kepler-vinoo-ganesh-0thtvuc) — Kepler independently reinforces Decagon’s discovery mechanism: small field fixes and repeated customer language become inputs to durable product capabilities.
- [How Forward Deployed Engineering is done at Ramp — Leo Mehr](https://feed7.dev/p/how-forward-deployed-engineering-is-done-at-ramp-leo-mehr-1w1rztz) — Ramp’s structured intake is a prerequisite for Decagon’s reuse decision because urgency, users, workarounds, and platforms help distinguish recurring needs from account-specific requests.
- [Forward Deployed Engineering at Cursor — Pauline Brunet](https://feed7.dev/p/forward-deployed-engineering-at-cursor-pauline-brunet-1wkd3r1) — Cursor’s customer ownership and documented handoff address Decagon’s risk that fast custom patches leave customers dependent on a brittle black box.
- [How Forward Deployed Engineering is done at Factory — Eno Reyes](https://feed7.dev/p/how-forward-deployed-engineering-is-done-at-factory-eno-reyes-0zgscmd) — Factory’s instrumentation and validators provide an implementation consequence for Decagon’s model: the path from field request to reusable feature needs observable outcomes and validation before autonomy expands.

## Context Map

- Layer: agent
- Domains: coding
- Topics: harness-engineering, context-engineering, enterprise

## Uncertainty

- The approach depends on distinguishing repeatable needs from genuinely customer-specific ones. Productizing too early can overgeneralize a single account’s constraints, while moving fast with patches can leave customers owning a brittle black box.

## Agent Instruction

Use this item as source-backed context. Do not invent claims beyond the linked source. If this item conflicts with another source, call out the conflict.