How Decagon Does Forward Deployed Engineering

These are my notes from the Forward Deployed Engineering workshop at @aiDotEngineer on June 30, 2026.

Session: How Forward Deployed Engineering Is Done at @DecagonAI
Speaker: Sunny Rekhi
Company: Decagon
X: @sunnyrekhi

Company Context

Decagon builds a 24/7 AI customer service agent.

It lands by deflecting complex support cases from human agents, then expands into proactive revenue generation once integrated.

It serves:

• Financial institutions.
• Retail and consumer brands.
• Large enterprises.
• Mid-market customers.

Two Types of FDE

Decagon described two FDE shapes:

• Agent configuration: training the AI brain on responses, tone, and actions.
• Enterprise problem-solving: frontline escalations and cross-account pattern recognition.

The FDE bar is identical to core engineering: same hiring bar, same reporting structure, same team.

Evolution at Scale

Early on, agent software engineers did everything with customers.

At 500 people, the function split into:

• Agent builders.
• Technical reviewers.

The scarce skill is now constraint and systems thinking, not coding.

Agents must be owned and understood by customers, not black-boxed.

Principles

• Scope success metrics upfront.
• Staff deals with vertical experts.
• Prove value on a narrow use case, then expand.
• Act as advisor, not executor.
• Make everything self-service over time.

My Take

Decagon's version of FDE is interesting because the bar is identical to core engineering. This is not a lower-status implementation role; it uses the same hiring bar, reporting structure, and team model.

The important scaling lesson is that coding becomes less scarce over time. Constraint thinking, systems thinking, and making customers real owners of their agents become the higher-leverage skills.