Over the past few months, a wave of AI vendors has rolled out “new” forward-deployed engineering (FDE) programs. Microsoft committed $2.5 billion and 6,000 people to a new unit called Microsoft Frontier Company, and AWS put $1 billion into its own FDE organization just weeks earlier. Both are pitched as accelerating AI adoption and surfacing new use cases inside customer environments, and neither is alone: OpenAI and Anthropic have made similar moves this year. I have sat through several of these announcements and pushed back with hard questions in vendor briefings. My take: the investment itself is worth making. But for most of these vendors, launching an FDE program right now is as much about fashion as it is about revenue.

FDE is not a new concept. Palantir Technologies pioneered it more than a decade ago, building it into a genuine early advantage in defense and intelligence work. Palantir’s own delivery model has changed since then, but the core tenets still hold. What is new is the attention: the term has only become an industry buzzphrase in the past two years, largely because Palantir’s recent AI and data-project wins have made “Forward Deployed Engineering” a label every vendor now wants attached to its own go-to-market motion.

Vendor-Sponsored Technical Education Is an Old Playbook

Vendors sharing technical resources with strategic customers is not new. It is how incumbents have primed revenue and captured product feedback through every major technology shift. So the concept behind FDE is not new, and based on my research, neither are most of the programs being announced under that name. At best, they are an evolution of work vendors were already doing. At worst, they are a rebrand. That distinction matters because the FDE model depends on deep business-domain expertise as much as engineering talent, and many of the new programs are thin on the business side.

No Vendor Is Replicating Palantir’s Original Model, Including Palantir

Palantir built its original FDE program as a bet on an underserved market, not as a sales tool. When the program launched, Palantir had no packaged products; its engineers had to figure out what the intelligence and defense communities actually needed and build bespoke solutions from scratch. That services-first approach eventually produced products, and once it did, Palantir’s own FDE motion shifted too, with engagements becoming shorter and more template-driven. Most of the new vendor programs, though, are built entirely around existing products from day one. That is a sales motion wearing the original model’s clothing, not a restatement of Palantir’s founding vision.

Forward-Deployed Engineering Has Already Split into Three Models

In recent research, I have identified three distinct versions of FDE now on the market:

Vendors and buyers should be explicit about which version they are getting, because the value, cost, and risk profile of each is different.

What Enterprises Should Demand Before Saying Yes

If your company is large or technically interesting, and you already have a solid relationship with an AI vendor, there is a good chance you will be offered an FDE engagement in the coming months, possibly at a discount or subsidized against future purchases. Treat it exactly like any other consulting or professional-services engagement: verify the team’s business-domain depth, not just its engineering talent, and do not assume the industry has settled on standards or best practices for these programs, because it has not. Vendors serious about this model need to staff it with as much business expertise as engineering headcount, or the “Forward Deployed” label is just marketing dressed up as strategy. And enterprises should go in assuming that some marketing support for the vendor is part of the price of admission. Right now, that is exactly what it is.