Forward deployed engineer vs. solutions engineer

The short answer

A solutions engineer (SE) sits in pre-sales. They run proofs of concept, map features to a pitch, and hand the account to customer success when the contract is signed. A forward deployed engineer sits in post-sales delivery. They take commit rights in the customer's repo, build against real data and real security constraints, and are measured by production — not by a demo that impresses a buying committee.

If your AI vendor's SE left and the workflow still is not in production, you do not need another POC. You need an embed. Start with what a forward deployed engineer is if the role is new to you.

Side by side

Why the SE model breaks on enterprise AI

Vendor SEs are good at showing what a model or platform can do on clean data. Enterprise AI fails on the opposite: entitlements, PII, SOAP APIs, change-control boards, and a claims system that has no documented schema.

The SE is not incented to live in that mess. Their quota is the signature. Once legal and security start the real review, the SE has moved to the next opportunity. That is why so many ChatGPT Enterprise and Databricks deals stall between "successful POC" and "running on production traffic."

OpenAI, Anthropic, Databricks, and Palantir all built forward deployed teams for this reason. The Palantir FDE role is the original version of that post-sales embed.

What an FDE does that an SE will not

• Opens PRs in your repo against your CI, not a vendor sandbox.
• Sits in your standups and takes tickets like a staff member.
• Designs retrieval, evals, and access control against your documents and your identity provider.
• Stays through security review instead of escalating it as "customer IT."
• Hands back a runbook and a trained owner. The SE hands back a CRM note.

Architecture calls such as RAG vs. fine-tuning get made on real data during the embed, not on the POC dataset.

When you still need a solutions engineer

Keep the SE motion when:

• You are evaluating a platform and need a fair POC before you buy.
• The question is "does this product fit" rather than "how do we ship our workflow."
• You already have an internal team that will own production after the deal.

Bring in an FDE when:

• The contract is signed and nothing is in production.
• The POC cannot be reproduced on real customer data inside the security perimeter.
• You need the IP in your repo, not another vendor environment.

A simple rule for CTOs

If the person cannot merge to main in your repository, they are not doing forward deployed work. They may still be valuable. They are just in a different job. Do not staff a production AI program with a pre-sales loop and hope it graduates.

Compare FDE vs staff augmentation if the alternative you are considering is contractors rather than a vendor SE.

FAQ

Can a solutions engineer become an FDE?

Some do. The jump is from demo-quality work to owning production under someone else's compliance regime. That is a different hiring bar — see the hiring guide.

Do FDE agencies replace the vendor SE?

No. The vendor SE helps you choose a platform. The FDE helps you ship a workflow on that platform, on your stack, with your data.

Is "forward deployed" just Palantir branding for professional services?

Palantir coined the title, but the job is specific: embed, commit rights, production outcome. Most professional-services benches do not work that way. The Palantir version is covered separately in Palantir forward deployed engineer.