Skip to content
All notes

AI delivery · 4 min

What a forward deployed engineer actually does

A forward deployed engineer sits with the customer, finds the problem that matters and ships the fix into their real systems. I have done a version of that job across 40+ client projects, and at a small business it looks different from the job ads.

A forward deployed engineer, or FDE, is an engineer whose desk is effectively at the customer. They learn how the customer really works, find the problem worth solving, build the solution against the customer's own data and tools, and stay until it runs. Palantir's job posts for the role describe it as understanding the customer's greatest pain points and designing end-to-end solutions for them.

I have done a version of the job for small and mid-sized businesses through my studio, across 40+ client projects, and on the five products I build and run. This is what the role looks like from inside that work, and where it differs from the way it is advertised.

What does a forward deployed engineer do?

  • Discovery at the customer. Sitting with the people who will use the thing, watching the work, and writing down what actually happens rather than what the process document says.
  • Building against reality. Their data, their logins, the spreadsheet that runs half the business, the old system nobody is allowed to switch off.
  • Shipping into production. The deliverable is a tool their team opens on Monday morning, with real users and real failures.
  • Staying through the first weeks of use, because that is when the real requirements show up.
  • Carrying what they learned back to the product team, so the next customer gets it built in.

Forward deployed engineer vs solutions architect

The two titles overlap and both get used loosely, so here is the line I would draw. A solutions architect designs how a product fits a customer and usually hands the build to someone else. A forward deployed engineer designs it, writes the code, wires the integrations and owns whether it works in production.

In AI work the gap between the two grows, because the hard part of an AI feature is rarely the model call. It is the data the model reads, the cases where it is confidently wrong, and what the business does when that happens. You only find those by running the thing on the customer's real inputs, which is the forward deployed half of the job.

What it looks like at a small business

Most ads for the role come from software companies selling to large customers, and the title has started to appear in Israeli job listings as well. At a business with ten or forty people the job has the same shape and different weight.

  • There is no IT department to hand anything to. Whatever I build, someone non-technical will own after I leave, so it has to be boring to run.
  • Discovery is short and direct. The owner is usually in the room, and the real process is whatever the two people who do the work every day actually do.
  • The budget decides the architecture. A small business will not pay to maintain a clever system, so the right answer is often a smaller tool on infrastructure they already pay for.
  • The person who asked for the AI feature is often the one most worried about it. Showing them where it can be wrong, before they find out alone, is part of the delivery.

What I check before I call it delivered

  • The team that will use it has used it on their own work, without me in the room.
  • Someone on their side knows what to do when it fails, and how they will notice that it failed.
  • Their data stays theirs: who can read it, where it is stored, and what happens to it if they stop paying for the tool.
  • The monthly cost is written down, including the AI usage, so nobody is surprised by the first invoice.

What is a fractional forward deployed engineer?

A fractional forward deployed engineer does the same job for part of the week, often for more than one company, instead of sitting full time inside a single customer. It suits a company that needs AI built into how it works but has neither the volume of work nor the budget for a full-time hire, which describes most small businesses.

  • The scope is one problem at a time, with a clear end: a working tool in production and someone on their side who owns it.
  • Discovery is compressed into a few sessions with the people who do the work, rather than weeks on site.
  • The handover starts on the first day, because the engineer will not be there every day after it ships.
  • It costs less than a full-time hire, and someone on their side already owns it when something unexpected happens.

Who needs one

A company needs forward deployed engineering when the gap between its product and a customer's reality is too wide for documentation to close. Small businesses meet the same gap from the other side: they bought tools, or had an AI app built, and it does not fit how they work. The fix in both cases is the same kind of person, someone who will sit with the work, write the code and stay until it runs.

If you are hiring for the role, ask candidates for something they shipped into a customer's real systems and what broke in the first two weeks. The answer to the second half tells you more than the first.

Add Ran Shoshan as a preferred source on Google

I read everything myself. If it is a fit I will say so fast, and if it is not, I will say that too.