Anthropic's Forward Deployed Engineering 101

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

Session: Forward Deployed Engineering 101 - @AnthropicAI
Speaker: Kevin Bai
Company: Anthropic
X: @zkevinbai

What FDE Is

FDE means selling outcomes, not software or services.

Engineers embed with customers, understand their business, and build solutions on the customer's platform. The customer buys a result, not a license or consulting hours.

Why FDE Exists

FDE fits a specific go-to-market gap: a technically complex product sold to a non-technical enterprise buyer.

If both the product and buyer are technical, self-service can work. If the product is simple or configurable, standard SaaS sales can work. A complex platform plus a non-technical enterprise buyer is where FDE matters.

Platform Requirement

FDE only works if engineers build on shared primitives, not from scratch.

Without shared primitives, every deployment becomes a unique codebase and maintenance becomes unsustainable.

FDEs assemble primitives into customer-specific solutions. They also scout what should be generalized back into the platform.

Why Now

AI makes customizable software easier to build.

Customers often do not know how to implement what they bought. FDE scales the design-partner motion used by early-stage B2B startups.

Should You Build FDE?

Ask "do I need one?" not "do I want one?"

The must-have condition is a technically complex product sold to a non-technical buyer. Also ask whether you have, or will build, a platform.

The ideal profile is a customer-facing software engineer.

My Take

Anthropic's version is the cleanest definition: FDE means selling outcomes, not software or services.

The constraint is just as important. FDE only works if the company has shared primitives to build from. Without a platform, every customer deployment becomes its own codebase, and the maintenance burden eventually overwhelms the motion.