Forward Deployed Engineering 101: The Origins, Essence, and AI-Era Revival of Palantir's FDE Model

Forward Deployed Engineering 101

Sending engineers to the customer site isn't about "consulting"; it's about **solving business problems for the customer**.

The core of FDE is "selling outcomes," not "selling software" or "selling people."

**Source**: These notes are based on Kevin's public talk at Anthropic, "Forward Deployed Engineering 101" (YouTube, 2026), about 17 minutes long (including Q&A). Kevin is a Member of Technical Staff on Anthropic's Applied AI team; previously he was an early Palantir employee and employee No. 1 on Rippling's FDE team, which he grew to about 25 people within a year.

Video link: https://youtu.be/KwhgfwOSToQ

1. The Core Thesis: FDE Solves Only One Specific Scenario

Kevin draws a 2×2 matrix:

FDE is the answer only when your product is highly complex and your buyer is completely non-technical.

Investment banks, oil and gas, government, large enterprises: they can't hire or retain deep engineering talent, but what they want is "10% more on the shelves" or "higher sales throughput." **They don't care how the data is organized, and they shouldn't have to.**

2. Palantir's Breakthrough: From "Selling Software" to "Selling Outcomes"

Two dead ends of a pure software business:

1. Explaining value is expensive: "I organized your data for you," but the business side isn't buying it
2. Customer success depends entirely on the customer: the customer not only pays but also has to train their own team before they can use it → a terrible business model

Palantir's solution: bundle the product and the service.

The customer isn't buying software, or someone's hours.

The customer is buying an **outcome**.

You send an engineer who knows the Foundry platform; they dig into the customer's business, figure out the problem, and then **build the solution directly** on Foundry.

Analogy: a waiter at a fine-dining restaurant. The customer doesn't need to understand the kitchen; the waiter "takes care of everything for you."

The data: Fortune 500 public SaaS companies ranked by ACV

• Palantir: $4M
• ServiceNow: $1.2M
• Workday: $600K
• None of the others has an ACV above $500K

The headcount is wild too: a company of a few thousand people reached an absurd valuation.

3. The Essence of FDE: Industrializing "Design Partnerships" at Enterprise Scale

FDE = Design Partnership × Enterprise Scale

Early-stage B2B startups find product-market fit through design partnerships:

• The founder works side by side with the customer
• The founder invests time, effort, technology, and resources
• The customer only provides context on the problem
• The founder delivers a solution that truly fits the need

Palantir's claim: who says design partnerships are only for early-stage startups? Just scale them to the enterprise.

But You Must Have a Platform

⚠️ FDE is not a dev shop

If every FDE writes code from scratch for customers:

• You end up with 55 repos nobody wants to take over
• Maintenance costs eat all your profit
• Your engineers quit

The defining trait of FDE: **always build on top of the platform; never write from-scratch code**.

The platform provides primitives, and FDEs assemble those primitives into workflows / apps / solutions.

4. Should You Build an FDE Team? Ask Yourself Two Questions First

Question 1: Is my business in the "selling a complex product to non-technical buyers" quadrant?

• If not → FDE isn't for you
• Complex product + technical buyer → DevRel
• Simple product + non-technical buyer → sales-led growth

Question 2: Do I have a platform (or am I willing to invest in one)?

• No platform → don't do it
• A platform with insufficient primitives → FDEs are your scouts
• A strong platform → go for it

Kevin's words: "I could not begin to stress the amount of maintenance burden that will be on your team even if you have a robust platform, never mind if you don't."

5. Why Everyone Is "Forced" to Become Palantir in the AI Era

The 2026 paradigm shift: almost every platform is becoming agentic = almost every platform is becoming customizable.

The consequences:

1. For every product you sell, customers don't really know what you're selling
2. Customer success depends entirely on the customer's ability to implement, which they simply don't have
3. Whether you're trying to upsell or land-and-expand, relying on customers to get started on their own doesn't work

Kevin's take: this isn't "everyone suddenly realized FDE is great and copied it." Rather, **the business itself has changed**: platforms becoming agentic = every platform needs an implementation partner, and FDE is the scaled version of an implementation partner.

6. Key Q&A

Q1: How "atomic" should primitives be?

A: It depends.

• One extreme: the app is 60% done and users customize only 40% → coarse primitives
• The other extreme: highly granular configuration for a broad range of customers → like AWS DynamoDB, which gives you the lowest-level puzzle pieces
• The deciding factor: how broad your user base is

Q2: How should multiple FDEs collaborate on the same project?

A: Strongly encouraged.

• It prevents a single point of failure: one person goes on vacation and the project stalls
• Cross-company collaboration is essentially a contractor relationship; be clear about who owns what

Q3: How do you split code written by FDEs vs. the platform team?

A: The golden rule:

• Customer-specific → stays on the FDE side
• Generalizable → eventually absorbed back into the platform

Early on the primitives aren't complete? That's fine. FDEs are "probes": you use them to discover what customers repeatedly need, which in turn defines what the platform should build next.

Q4: What does the perfect FDE look like?

A:

"FDE is nothing more than a customer-facing software engineer."

• You'd hire them as a software engineer (technically strong)
• You'd also put them in front of customers (strong in communication + business)
• The rest you have to figure out through practice

References

• Original video: Kevin, *Forward Deployed Engineering 101*, YouTube, 2026.

https://youtu.be/KwhgfwOSToQ

• Speaker: Anthropic Applied AI · Member of Technical Staff; former FDE at Palantir / Rippling

These notes were compiled by Hermes Agent from the video transcript. Subjective content such as "further thoughts for the author" was removed; only the speaker's key ideas, frameworks, and Q&A are kept.