Forward-deployed engineering is a business model in which a software company sends its own engineers to work inside the customer, connecting the software to the customer's data and building the first real uses, often before a long contract is signed. The labor up front is the cost of the sale; the payoff is an account that keeps expanding.
RewiredNot new, but AI made it the way enterprise AI gets sold: a general model does little inside a company until someone connects it to the data and the work.
How it makes money
Most software companies sell a product and leave the hard part, making it work in a particular company, to the customer or a consulting partner. Forward-deployed engineering takes the hard part in house. The vendor's own engineers go to the customer, wire the software into its data and build the first uses that matter.
The economics are those of an investment. Labor is spent before revenue arrives, and recovered as the account expands to new teams and new uses. In business model terms, the vendor changes who does the work, and bets that doing it makes the customer stay.
One company, read closely
Palantir's 10-K describes the cost without varnish. The company says it often launches pilot deployments "at our expense" without a long-term contract in place, and that its sales model "has historically required us to spend months and invest significant resources working with customers on pilot deployments at no or low cost to them." Its answer has been to compress the work: AIP bootcamps, it says, "allow us to deliver real workflows on actual customer data in days."
The payoff shows in how accounts grow. Average revenue from Palantir's top twenty customers was $93.9 million in the twelve months to December 31, 2025, up 45% from $64.6 million a year earlier. Revenue from US commercial customers rose 109%, to $1.5 billion from $702.3 million. And the contracts follow the work: at the end of 2025, Palantir counted $11.2 billion of remaining deal value, $6.8 billion of it with commercial customers.
How it fails
The filing names the risk in the same breath as the method: there is "no guarantee that we will be able to convert customers from these short-term pilot deployments to longer-term revenue-generating contracts." Some pilots, it says, have produced little or no revenue at all. Every pilot that ends there is engineering time given away.
Our read
Forward-deployed engineering treats the first months of an account as an investment rather than a cost of sale. It works when the software, once running, keeps finding new uses inside the same customer, so that the engineers' first project pays for the ones that follow. That is what Palantir's top-customer figures suggest happened there.
AI makes the pattern more common because a general model does little inside a company on its own; someone has to connect it to the data and the work. If you buy AI this way, ask who owns what the vendor's engineers build, and what it costs to keep running once they leave. If you sell it, judge each pilot by how quickly it turns into a second use, because that is where the model pays.
Questions
- What is a forward-deployed engineer?
- An engineer employed by a software vendor who works inside a customer's organization, connecting the vendor's software to the customer's data and building working uses with the customer's teams.
- Why do software companies send engineers to customers?
- Because complex software, AI in particular, does little until it is connected to a company's own data and processes. Doing that work for the customer shortens the time to a working result and makes the account more likely to expand.
- What is an AIP bootcamp?
- Palantir's name for short, hands-on sessions in which it builds working AI uses on a prospective customer's own data. The company says they deliver real workflows on actual customer data in days.
Sources
Every figure and quotation on this page traces to these filings. “Our read” is Stratrix’s judgment, drawn from them. Analysis and data, not advice.
More from Stratrix
“What is stopping a working pilot from reaching production?”
See the pack — $499 →The Business Model Wizard
Make the seven choices that define your model, test your numbers against real companies, and meet the rival AI would build against you.
Free · about twelve minutesBuild your Model Thesis →Test a decision, free
Set a live decision against the companies that made it before you.
AcquisitionsThe Distance TestDivestituresThe Divestiture TestDiversificationThe Diversification TestMake or buyThe Make-or-Buy TestWeekly email
Stratonomics, every Tuesday
One data story every Tuesday: what happened, what it means, and the figures you can quote. Free.