The newspaper of the Forward Deployed Engineer
In-depth pieces for working FDEs: original data, primary sources, lessons you can use.
No company has published this figure. The playbooks still agree on one point: how many accounts you can carry depends on the phase each account is in and on how much of the work has been turned into tools.
09/10/2026
Salesforce has written into its FDE job listing that code may come from AI tools, but the person accountable for the result must still be an engineer.
Many people now treat FDEs as a cure-all. Phosai Labs, a services and consulting firm, offers a precise hiring signal and a 20% threshold for spotting when an FDE team has drifted into sales support.
Y Combinator tells founders to work like engineers embedded with customers. The reverse is only half true, and the missing half decides whether you build a product company or a services business.
Microsoft lists Kubeflow as an equivalent of its own product, and Google runs KFP code directly. A feature comparison no longer helps an FDE pick a platform.
One rule wants your system to remember for a long time. Another wants it to forget quickly. Both land in the same place: the log table you write today.
A feature matrix tells you what each tool can do. For a forward deployed engineer, two other questions matter more: where the client's data is allowed to go, and what the client actually needs to find out.
Even Anthropic advises starting without a framework. So the question at a client site is no longer which library to use, but which part of the work you are willing to hand to someone else's library.
AI companies have committed about $10bn in 12 months to building FDE teams. Whether the model survives depends less on how many people they hire than on what those people leave behind.
The distance between "1+ years of experience" in a Palantir posting and "2+ years managing FDEs" in an OpenAI one says more about the profession than any salary report.
