The Dirty Secret of Forward Deployed Engineering — Natalie Meurer, Sierra

Summary of Technological Concepts, Product Features, and Analysis

What “forward deployed engineering” is (and why the term is controversial)

• The speaker (Natalie Meurer, Sierra) argues that the “dirty secret” is that forward deployed engineering (FDE) doesn’t exist as a coherent, stable definition—the term has been stretched to cover many different roles.
• Despite this ambiguity, she frames FDE as one of the hottest job tracks in AI, with meaningful continuity across different eras.

Historical evolution (Palantir → modern practice)

The talk describes FDE as a timeline of expanding responsibilities, rather than discrete phases:

1. 2008–2012: FDE resembles DevOps / platform stability

   • Early deployments focused on keeping platforms stable in customer environments, including on-prem and cloud instance management.
   • Example: her onboarding project at Palantir involved deploying software to an EC2 instance, where issues could be urgent (e.g., customer reports something unplugged at 2:00 a.m.).

2. 2012–2016: Data integration becomes central

   • Palantir develops data integration software, so FDE must help customers integrate data across many systems.
   • Engineers model customer data using an “ontology” concept (Palantir terminology), essentially a taxonomy for data.
   • Implementations often relied on older tech stacks (she mentions Java / Java Swing).

3. 2016: Custom solutions and dashboard builders (Slate → dashboards that write back)

   • FDE helps build custom solutions for known problems.
   • Mentions Slate, a drag-and-drop builder mapping UI components to data sources (conceptually compared to tools like Retool).
   • Emphasizes dashboards must write back to data sources; otherwise they decay and become less useful.
   • This connects to Foundry, Palantir’s main platform, focused on moving data toward decision-making.

4. 2020 onward: Enablement + scaling customer self-sufficiency

   • Palantir shifts toward empowering customers to do more themselves (less dependence on shipping engineers onsite).
   • Example: Airbus Skywise enabling thousands of engineers to use the Foundry platform.
   • Mentions an AIFTE-style system using large language models to assist similar work.
   • FDE also includes training/bootcamps (she references AIP bootcamp-style enablement).

Core continuity: “accountable to the customer”

Across DevOps, data integration, custom solutions, and enablement, the key thread is:

• Every deployed engineer is accountable to the customer for outcomes in some form.

This accountability becomes the lens for why FDE matters.

How FDE changes when code becomes cheaper (agents & product impact)

• With AI agents, code generation becomes faster and cheaper (e.g., “prompt → agent → good output”).
• The speaker argues FDE engineers can now do more than prototype:
  • They can build end-to-end solutions using coding agents.
  • They can translate customer-accountability signals into product improvements—such as stability improvements or better data integration.

Blurring boundaries: product engineering converges with customer-facing work

She claims role boundaries are blurring:

• Product engineering becomes more client-facing
• A strong product/infrastructure engineer should also understand deployment (DevOps) and customer needs
• FDE “gets larger” by stacking skills rather than remaining a single role

Pricing model analysis: toward outcome-based pricing

She connects FDE with changing approaches to software monetization:

• Sierra uses outcome-based pricing: pay based on value delivered.
• Contrast with older models:
  • Seat-based pricing: harder to tie to measurable outcomes
  • Usage-based pricing: common with foundation models (e.g., OpenAI/Anthropic), but outcomes are hard to attribute directly
• With AI agents, she predicts pricing will shift toward outcome-based because it’s outcomes (sales, solved inquiries, achieved customer goals) that matter most.

“Agent engineering” vs “FDE” (discipline stacking and renaming)

Earlier, while at Sierra, she wrote:

• Agent engineering should be treated as a subdiscipline of AI engineering
• It should have customer accountability similar to FDE, but remain a distinct domain

Now she revises the framing:

• Everything trends toward being forward deployed engineering, including:
  • product engineering
  • solutions engineering
  • customer engineering / enablement
  • building and deploying agents

Closing stance / hiring note

• The slogan: “Forward deployed engineering is dead. Long live forward deployed engineering.”
  • Meaning: the term is incoherent, but the underlying customer-outcome practice is expanding.
• Sierra is hiring across forward deployed engineering roles.

Main speakers / sources

• Speaker: Natalie Meurer (Head of Agent Engineering, Sierra)
• Sources mentioned:
  • Palantir (historical reference and platform concepts like Foundry, Slate, “ontology”)
  • Emergence Capital (referenced in discussion of code becoming cheap / market shift)
  • Elliot Greenwald (Sierra head of go-to-market ops; mentioned via a blog post reference)
• Companies/figures mentioned:
  • Google (GCP)
  • OpenAI
  • Anthropic
  • Aaron Levie
  • Airbus Skywise
  • AIFTE (Sierra internal/follow-on work mentioned by name)