AI Models Are The Gold, Forward-Deployed Engineers Are The Gold Miners

People, not models, build the moat. In 2023, every enterprise leader suddenly had the same thought: We need to figure out our AI strategy. In 2024, many raced to adopt the newest models. But in 2025, the most forward-looking companies are asking a different question: How do we implement AI safely, reliably, and effectively? 

As AI becomes more widespread, with adoption surging even as leading labs like OpenAI and Anthropic maintain an edge over open-source models, the true differentiator isn’t which model you choose. It’s how you’ve implemented AI that determines whether your AI systems are actually solving real business problems. And the most in-demand role helping companies accomplish this is the Forward Deployed Engineer (FDE).

FDEs are engineers who are embedded directly with customers or internal business units to understand and automate bespoke workflows, all while building trust with customers. Unlike traditional product engineers or consultants, they’re close enough to the end user to build context and technical enough to turn that context into working software. Palantir pioneered this approach with their clients, and the model is rapidly spreading to AI-native startups.

At a time when just 1% of companies have reached full AI maturity, successful implementations are becoming a bottleneck, leading demand for FDE talent to skyrocket. Across our Emergence portfolio, we’re seeing a surge in companies hiring or reshaping engineering roles to fit this mold, including leaders like Bland and Federato. The rise of FDEs is not a temporary trend. They represent a new way to deliver enterprise software, where the value is tied not just to features, but to how well you solve a user’s business problem.

The key insight from Palantir’s model is that selling software isn’t enough. If the buyer can’t drive meaningful usage or solve real operational problems, your software won’t stick. FDEs bridge this gap, often by rolling up their sleeves to do the work themselves, customizing implementation around a clear ROI. They aren’t just a layer in the AI stack. They’re a new path to product-market fit.

SaaS was about digitizing business processes. This new era of software is about using agentic AI to rethink workflows and automate them. FDEs play a key role in helping navigate this shift to fully harness AI models. AI models are the gold, and the FDEs are the miners. Without miners, you can’t get the gold out. 

SOPs Are a False Security Blanket

Most companies believe they’ve already documented their critical business logic in standard operating procedures or SOPs. But SOPs are corporate fiction: static, incomplete, and often wildly outdated. AI systems need behavioral fidelity—how the work actually gets done—not the idealized version that’s written in the wiki.

74% of companies struggle to scale the value of their AI efforts, often because their automation initiatives are based on flawed assumptions and rigid documentation. FDEs help close that gap in three ways:

1. Discovery: They shadow users and reconstruct how processes truly function, not just how they’re supposed to.
   
2. Integration: They execute the last-mile work of wiring AI into real tools like ERPs, CRMs, and call center platforms.
3. Trust: They build relationships with the people doing the work, which helps drive adoption and reveal nuance.

Understanding process fidelity is now a prerequisite for AI-driven transformation. FDEs are the human interface that makes that possible. And as AI increasingly automates human labor (not just digital workflows) demand for this function is exploding. FDEs are the frontline force ensuring these systems actually work in production.

The Internal Scale AI

We already accept that AI models require labeled training data. Companies like Scale AI built billion-dollar businesses labeling images and documents for general-purpose models. FDEs perform the same function for your internal systems: they “label” workflows through observation, contextualization, and code.

This step is essential to building reliable, usable AI applications tailored to your business. If you skip this layer, you’re likely to end up with brittle, untrusted software that falls short of expectations.

And unlike generic model trainers, FDEs help you retain and encode your proprietary logic. They are your internal Scale AI.

The Software Blind Spot: Enabling the Gold Miners

The SaaS ecosystem has built an entire arsenal of tooling for go-to-market teams, product teams, and developers. But there’s almost nothing purpose-built for the people doing some of the most sensitive, strategic work: FDEs. If FDEs are the gold miners, they’ll need picks and shovels to get the gold out. 

That said, the role of the FDE isn’t something that should or can be automated away. At early and growth-stage companies, the work is highly bespoke. Each engagement reveals new friction, new systems, and new user realities. That’s the point.

But that doesn’t mean we can’t improve leverage. The real opportunity is enabling FDEs to spend more time on customer discovery and implementation, and less time on repetitive work. Tools that assist with things like conducting and synthesizing customer interviews, or mapping process logic can free up FDEs to focus on high-trust, high-impact work.

The FDE picks and shovels are coming. And the companies that build or adopt it early will gain leverage fast.

The Next Phase of Enterprise AI

We often talk about AI as if the future will be run by a single, monolithic model. But enterprise AI isn’t a static deployment. It’s a continuous loop between humans and systems.

Agentic software will evolve with people in the loop. The applications companies use will be finely tuned to their internal systems and workflows. That evolution will be guided not by prompts or dashboards, but by embedded engineers—FDEs—who understand the terrain.

AI outputs precision. But people input nuance. FDEs ensure those two sides stay in sync.

The Real Moat Is Human-Led, Not Model-Driven

AI won’t be self-sufficient. Not now. Not ever. The companies that win will be the ones that invest in human-in-the-loop expertise, and treat their operations not just as things to automate but as assets to protect.

FDEs aren’t just translators. They are miners of real-world process knowledge, turning client context into code, customizing products to meet business needs, and making AI platforms more usable, reliable, and sticky. In doing so, they create a new kind of moat: one built not from proprietary models, but from proprietary implementation.

If you’re not empowering them, you’re missing your biggest leverage point.