Service-as-Software needs forward deployed engineers
If you sell finished work, someone has to make it work inside the client's company. That someone is a forward deployed engineer. It is the most expensive part of the model, and the one that decides whether it pays.
What a forward deployed engineer is
The term comes from Palantir. Its product engineers build the same feature for many customers. Its forward deployed engineers, whom it calls Deltas, do the opposite: they work for a single customer and build everything that customer needs.
They work with the client's team, learn how the work is really done and make the software do it. Palantir judges them by the client's result, not by the code they ship.
A forward deployed engineer is not a consultant. A consultant recommends. A forward deployed engineer builds, ships and stays until it works.
Why selling finished work needs someone inside
A tool can be sold and left on a shelf. Making it work is the buyer's job. Finished work cannot be left anywhere: if it does not arrive, there is nothing to bill.
And the work lives in the client's systems. Their CRM, their spreadsheets, the exceptions nobody wrote down. No AI model knows them until someone goes and looks.
That is why the companies that sell AI agents are hiring these engineers. In June 2025 a16z called it the hottest job in startups.
Where it breaks
In January 2026 Marc Andrusko, a partner at a16z, sounded the warning in “The Palantirization of everything”.
Copy only the engineers on site and you get many custom builds that nobody can maintain or upgrade. In his words, an “Accenture for X” with a nicer front end.
Most problems do not justify months of engineering on site. And people who understand both the business and the code are scarce.
His advice is to treat these engineers as scaffolding: work on site with an end date, a limit on headcount and pieces that can be reused.
How I apply it
I work as a forward deployed engineer: inside your systems, with your real data and with your team. It is how I already work inside Sesame HR, where each team is an internal client. With three limits from the first day.
The work on site has an end date. The diagnosis picks one task, and the build phase ends when the system meets the agreed standard. After that I maintain it, with no more building.
What is not specific to you is built to be reused: the checks and the monthly report. Only what is truly yours is made to measure.
The cost of each task is measured from day one. That number shows when engineering stops being the main cost, and whether the service is worth it.
What to ask a provider
Three questions before you sign.
Who will work with my team, and for how long?
What is built only for me, and what is reused?
When does the build phase end and maintenance start?
If the answer is “as long as it takes”, you are buying hours, not results.
Read next
- How I work
The process step by step, what you pay for and which work fits.
- Copilot or autopilot: where the money is and where the margin is
My essay on Sequoia's thesis and how to pick the first task.
- a16z: The Palantirization of everything
Marc Andrusko's essay, January 2026.
- Palantir: Dev versus Delta
How Palantir describes its forward deployed engineers.
Engineering with an end date
Write to me with a task you already pay someone else to do. I will tell you if it fits and when the build phase would end.
Tell me about the task