- Didier's Newsletter
- Posts
- The rise of forward-deployed engineering in finance
The rise of forward-deployed engineering in finance
How Palantir's playbook became the blueprint for agentic finance
Didier Lopes
November 25, 2025
A few weeks ago, we were at the HQ of one of our largest clients (close to 70bn AUM) for a week and although I had this post in draft for a couple of months, these words resonate even more with me now as we experienced this FDE work first-hand.
Anyway, let’s talk about FDE.
Let me start by sharing an excerpt by Palantir’s CTO on the origin of the name “Forward Deployed Engineer” and how it relates to how French restaurants works.
“In 2006, Alex Karp asked me if I knew why French restaurants were so good. I had no idea. He told me that at a French restaurant, the wait staff is actually part of the kitchen staff. They intimately understand the food, the methodology, and the technique. They are not merely carrying the food from the kitchen to the table, but are instead part of a subtle and complex system that affects kitchen operations. He wanted me to build that, but for engineering.
I did just that, and in 2007, I gave it the name “forward deployed engineering” (FDE) in homage to our customers. FDEs embed alongside our customers and work to ensure our software solves their problem and not some proxy for their problem. They’re crazy enough to get on a last-minute plane to Iraq, they’re smart enough to ship quality, same-day code, and their EQ is still high enough to talk to users (and maybe even enjoy it). Investors ridiculed us for creating a “services” role that would only serve to depress the margins of a software company. We didn’t look like the other SaaS businesses, something they eventually realized was a feature, not a bug. Just like French restaurants don’t blindly hand off food from an uninformed waiter to the diner, we didn’t believe in throwing our software over the wall in the hopes the customer would divine the correct meaning from it. This approach built an engineering organization with unmatched creativity, responsiveness, and focus on the primacy of winning.”
The quiet truth
AI-native platforms promise to "automate financial workflows". But enterprise wins aren't self-serve trials, they're proof-of-concepts plus forward-deployed engineers.
Engineers sitting inside client offices. Writing code on their networks. Building bespoke workflows that prove the platform works.
For every "AI assistant for analysts", there are FDEs buried in a buy-side firm's infrastructure, understanding data pipelines, writing agentic workflows, fixing edge cases, etc… The companies landing the biggest customers aren't winning on product alone, they're winning on proximity.
In finance, no abstraction replaces trust. This isn't a workaround. It's how intelligence software gets deployed in high-stakes domains.
Why forward deployment works
AI-native software isn't plug-and-play. Output quality depends on data specificity and governance. Every firm's architecture is different. Forward deployment brings engineering to the data.
An embedded engineer does three things pure software can't:
Translates domain into system. They learn how the firm defines "exposure" or "active position" and encode those meanings into their workflows.
Builds trust loops. Same Slack channels, real workflows, outputs validated against the firm's sources of truth.
Turns deployments into reusable abstractions. Each new workflow becomes productized logic that can be leveraged for the next client.
Most AI companies say they're selling intelligence. They're selling trust. Clients ask what peers are doing. They want the playbook, not just the platform. An engineer inside the firm, fixing pipelines and aligning definitions, builds credibility no deck can.
Trust compounds. Every successful deployment becomes a reference pattern the engineer can leverage for next client, obviously without sharing the actual workflow.
3. The Palantir playbook
The breakthrough wasn't "work closely with clients". It was turning embedded work into product leverage.
Each deployment as live R&D. Engineers discovered reusable patterns: data models, workflows, connectors, ontologies. Those patterns moved into the platform, so the next deployment started smarter.
The self-scaling loop:
Engineers deploy → deployment creates insight → insight becomes reusable → product gets stronger → next deployment accelerates
What looked like high-touch work was the most scalable form of enterprise learning. Agentic finance is doing this now. Teams embedding engineers aren't slowing product - they're accelerating it.
This is the playbook for the agentic era: scale product by deploying where it's most complex. Learn fast, abstract faster.
“(…) Specifically, the firm [PLTR] provides a modern AI data platform (i.e., analytical tools, data infrastructure software, AI tooling) that helps government agencies and enterprises make data-driven decisions by detecting unusual or previously undetectable patterns in large complex datasets. Its AI data platform is differentiated by its Ontology, which presents data to users in their own everyday terms and represents decisions in an enterprise, thus becoming a powerful tool for AI-driven decision-making. On top of this, Palantir deploys an effective and super-efficient go-to-market approach that attracts customers and quickly converts them into paying customers of its complex products. (…)”
4. The hybrid flywheel
PLG dominated the last decade because products were simple and users empowered to self-serve. AI-native finance breaks that logic, it's probabilistic, data-hungry, unpredictable in the wild.
PLG still matters as half the story. Bottom-up motion builds credibility and reach. But PLG alone doesn't close institutional deals or integrate with legacy systems. In finance, analysts often lack the time or mandate to tinker.
Forward deployment closes those gaps, but can't scale in isolation. Without strong product underneath, every engagement risks becoming bespoke.
The future isn't product-led or sales-led. It's deployment-led growth, powered by a product that learns faster with every field engagement.
Here's the loop:
Open distribution → rapid experimentation. Free tiers, open examples, extensible interfaces.
FDEs → institutional activation. Embedded engineers integrate, secure, productionize.
Field learning → product leverage. With every integration there are new teachings, that are folded back so the next deployment is faster.
Faster deployments → broader adoption. As the product gets smarter, less custom work is needed. The loop accelerates.
Community fuels product. Product fuels deployment. Deployment fuels product again.
One motion
Bottom-up and top-down are converging.
The open community and the forward-deployed engineer are two sides of one motion - one learns from scale, the other from proximity. Both feed the same outcome: a smarter, faster, more resilient product.
The field becomes part of R&D. Every deployed engineer is a node in the product feedback network. Every community user at the edge generates signal for where to go next.
The AI platforms that win won't just have the best model or most data, they'll have mastered this hybrid motion of openness and proximity.
The companies that merge field intimacy with open distribution will own the next era of financial software.
Proximity drives trust. Openness drives scale.
This is a great video from YC on FDEs: