The FDE Bubble: Enterprise AI's Costly Game of Pretend
Enterprise AI·October 21, 2026

Many enterprise AI startups pitch software that runs itself. Behind a number of them, though, sits a large team of forward deployed engineers, known as FDEs, who travel to customer sites, connect the product to existing systems, and fix things when they break. In a guest commentary, Anurag Gurtu, co-founder and CEO of Airrived, argues that this work looks much more like consulting or staffing than software, and that the industry has an incentive to present it otherwise.
The core problem, in his account, is how the money gets counted. When every new customer needs engineers on the ground, the income may still be reported as recurring subscription revenue. That makes the business appear to have software economics, when it is really carrying the cost of a services operation. The gap can inflate annual recurring revenue and, with it, the valuation investors are willing to pay. Gurtu also contends that many of these products fall short of the autonomy they advertise, since each new environment still needs significant human effort to get working.
For buyers, he suggests some practical questions before signing. How many engineers does a typical deployment take? How long does it last? How much of the contract value goes to people rather than the product itself? For investors, the telling figure is gross margin. A true software company generally keeps most of each revenue dollar after hosting and support costs. A company whose deals depend on large engineering teams may look far less profitable once that staffing is counted.
The broader point is less about the FDE model itself than about how it gets described. Hands-on implementation can be a legitimate business, and some customers want that help. The objection is to calling a services-heavy company a software company. Gurtu's argument, in effect, is that such a business should be judged and valued like the service it is, and that a company selling software should be able to show the product doing the work without a large crew in the room.
Reporting based on an external source.