The FDE Bubble: Silicon Valley’s Most Expensive Game Of Pretend

We have entered the era of the Forward Deployed Engineer (FDE) bubble, and it looks a lot like the old era of consulting firms wearing masks. In the rush to capture the enterprise AI market, the line between software-as-a-service and labor-as-a-service has blurred to the point of disappearing. If you are a founder or an investor today, you need to understand why this matters before the margins catch up to the hype.

The Great Software Illusion

For decades, the dream of Silicon Valley was the high-margin software business. You write code once, and you sell it a thousand times. The marginal cost of adding a new customer is near zero. That is why software companies trade at massive multiples compared to service businesses. But building autonomous agents that actually work in the messy reality of the enterprise is hard. It is so hard, in fact, that many of today’s hottest startups have resorted to a clever form of cheating.

Instead of building a product that works out of the box, these companies are deploying small armies of FDEs. These are highly paid, highly skilled engineers who move into a client’s office to manually bridge the gap between what the AI was promised to do and what it actually can do. On the balance sheet, this revenue is often booked as recurring software revenue. In reality, it is a professional services contract dressed in a hoodie.

The Problem with Human-Powered Autonomy

The term agentic AI implies a level of independence. An agent should be able to reason, plan, and execute tasks without a human holding its hand. However, the current enterprise landscape is littered with agents that require constant babysitting. When a startup sends ten engineers to manage a single deployment, they are not selling an agent; they are selling a temporary staffing solution.

This creates a dangerous feedback loop. Investors see the high revenue growth and pump in more capital at astronomical valuations. The founders, under pressure to justify those valuations, hire more engineers to handle more deployments. The gross margins, which should be 80% or 90% in a healthy software company, start to look more like the 30% or 40% margins of a traditional IT consultancy. The moment the hiring stops, the product stops working, and the growth plateaus.

Why Builders Should Care

If you are building in the AI space, the FDE model is a trap. It is tempting to use human labor to solve technical hurdles because it gets you to a signed contract faster. But you are building technical debt that is paid in human salaries rather than code refactoring. You are essentially building a business that does not scale.

True innovation in agentic AI will come from those who focus on the architecture of autonomy. This means building systems that can handle edge cases without a human engineer needing to write a custom script for every new client. It means focusing on reliability and self-healing systems rather than just flashy demos that require a support team to keep running behind the scenes.

The Investors Reality Check

For those on the funding side, the due diligence process needs to evolve. Looking at Annual Recurring Revenue (ARR) is no longer enough. You have to look at the Cost of Goods Sold (COGS). If a company’s headcount is growing in direct proportion to its customer count, you are not looking at a software company. You are looking at a services firm that is going to have a very difficult time scaling its way to a liquidity event.

We have seen this movie before with the first wave of AI startups that relied on manual data labeling. The companies that survived were the ones that eventually automated the labeling process. The companies that failed were the ones that remained dependent on human labor while trying to convince the market they were pure-play software companies.

The Path Forward

The bubble will eventually pop when the enterprise buyers realize they are paying software prices for what is essentially a managed service. When the FDEs leave the building, the software needs to stay behind and actually perform. If it doesn’t, the contract won’t be renewed, and the valuation will crater.

The winners in the next phase of AI will be the founders who are honest about what their agents can do. They will be the ones who prioritize building robust, autonomous cores rather than inflating their numbers with engineering-heavy deployments. It is better to have ten customers and a product that scales than a hundred customers and an army of engineers you can't afford to keep.

Takeaway for Founders

Focus on your gross margins from day one. If your product requires a human to survive, you haven’t built an agent; you’ve built a job. In the long run, the market rewards scalability, not just revenue. Don’t get caught in the FDE trap just to hit a quarterly target.

Read the original at Crunchbase News →