Forward Deployed Engineers: the builder and the bulldozer

The forward deployed engineer role is at peak zeitgeist right now.

This week Amazon committed $1B to an FDE org. OpenAI and Anthropic already did versions of the same thing. Google is hiring hundreds. Every VC-backed company looking for customer-facing engineers are hiring FDEs en masse.

And while the title has lost all meaning (from its humble beginnings at Palantir) & turned into a meme in nerdy tech circles, it's worth understanding if the hype is earned.

@levie has said FDEs one of the most in-demand jobs in tech, because deploying agents is far more technical than people realize. Software companies look more like service firms than ever before in order to effectively diffuse AI into a complex latticework, people, process, and system integration.

@emollick says they may disappoint, because the real problem isn't technical. It's rethinking the expertise & structure of your organization around AI.

I think they're both right, and the disagreement is the whole point.

The work is actually two different jobs. Call them the builder and the bulldozer.

The builder is the FDE. You cannot ship a good agentic solution without someone who's customer-facing, deeply understands the innards of an organization, shadows the people actually doing the work, process-maps every step, and takes the solution through the ringer before, while, and after building it.

The bulldozer is the one who clears the path for adoption. Different role entirely. It's an AI/Deployment Strategist, PM, or operational leader with the political capital to persuade the bosses, whose whole job is getting people to behave in the new way the rebuilt process needs. Buy-in from the top. Alignment from the ones doing the work. And enough aperture across the org to know what's needed to make reinvention stick.

Here's a live example.

We're building an agentic tool for a company that staffs hourly retail workers across a lot of stores. It checks that each worker is doing their in-store steps correctly, so you get company-level quality control over a distributed workforce.

The builder's (FDE) job is to shadow those workers in-store and rebuild the process to speed up every step without changing how the worker behaves.

But a perfect tool drives zero value if the workers don't use it. And getting them to use it comes down to incentives, training & enablement, and what their managers police. None of that is the engineer's job.

So I agree with Aaron & Ethan. FDEs are exceptionally important, and will be for years as the AI industrial complex takes hold in enterprises. AND they are but just one key ingredient as organizations try to shift their AI story from experimentation to measurable gains.