What does a forward-deployed engineer do at an AI startup?

Gets the product working in each customer's systems: integrations, data pipelines, customer-specific workflows, evals and the production rollout. They work under the startup's brand, and what each deployment teaches goes back into the product, so the next one needs less custom work.

Should an AI startup's founders do forward-deployed work themselves?

Early on, yes: the first deployments show the founders what the product is missing. It stops making sense when deployment work keeps the founders and core engineers off the roadmap, which is the signal to bring in FDEs.

Does an AI startup need an FDE or a solutions engineer?

It depends on where deals get stuck. A solutions engineer helps win the deal: demos, proofs of concept and the technical case before signature. An FDE delivers it: production code in the customer's systems after signature, and accountability for whether it works.

Is FDE as a service the same as outsourcing?

No. Outsourcing hands a scoped project to another team. FDE as a service places engineers inside the company's own team, in its repositories and under its brand, accountable for its customers' deployments, with what they learn going back into the product.

FDE as a service / guides

When does an AI startup need forward-deployed engineers?

By Anusheel Bhushan, co-founder of Plank

An AI startup needs forward-deployed engineers when the product works in the demo but each customer needs engineering to get it running in their systems, and that work is pulling the core team off the roadmap. Until then, the bottleneck is the product, and product engineers are the better hire.

Signs deployment is the bottleneck

You need FDEs when

Deals close, but deployments stall or never go live.
Every new customer needs integration work: their CRM, their warehouse, their auth, their data model.
Pilots succeed and still do not turn into production contracts.
Core engineers spend their week on customer calls and one-off fixes instead of the roadmap.
Enterprise customers ask to run the product in their own cloud, under their own security review.

You don't need them yet when

Customers set the product up themselves and succeed without your engineers.
You have no paying customers and are still finding product-market fit.
The gap is in winning deals rather than delivering them. That calls for a solutions engineer.
The task is beyond what current models do reliably. That is a research problem.
One customer needs one integration, once. A contractor or an implementation engineer will do.

What an FDE does at an AI startup

Forward-deployed engineers work inside the startup's team on its customers' deployments: integrations with each customer's systems of record, mapping their data into the product, the customer-specific workflows the product needs, evals on the customer's data, and the production rollout. What each deployment teaches goes back to the product team, so the next one needs less custom work.

In Plank's census of 904 FDE job descriptions, 51% ask for API integration work and 27% for data pipelines.

What employers ask of FDEs, from the census

Questions to ask a provider

How are your engineers trained for deployment work? If a provider cannot describe it, expect staff augmentation.
Will they work in our repositories, join our standups, and represent us to customers?
Who owns the code and the IP, and is that signed before any code is written?
How does what they learn reach our team, and by when?
What happens to a deployment when the engagement ends?

Hire in-house or use a service?

Hire in-house when forward deployment is a permanent part of how you sell and the domain knowledge compounds over years. Use a service when signed customers are waiting now, when demand comes in waves, or when the role is too new to interview for well. Many startups do both: a service carries the load while the first in-house FDEs are hired, with the handover planned from the start.

In-house FDEs vs FDE as a service, compared

Common questions

What does a forward-deployed engineer do at an AI startup?: Gets the product working in each customer's systems: integrations, data pipelines, customer-specific workflows, evals and the production rollout. They work under the startup's brand, and what each deployment teaches goes back into the product, so the next one needs less custom work.
Should an AI startup's founders do forward-deployed work themselves?: Early on, yes: the first deployments show the founders what the product is missing. It stops making sense when deployment work keeps the founders and core engineers off the roadmap, which is the signal to bring in FDEs.
Does an AI startup need an FDE or a solutions engineer?: It depends on where deals get stuck. A solutions engineer helps win the deal: demos, proofs of concept and the technical case before signature. An FDE delivers it: production code in the customer's systems after signature, and accountability for whether it works.
Is FDE as a service the same as outsourcing?: No. Outsourcing hands a scoped project to another team. FDE as a service places engineers inside the company's own team, in its repositories and under its brand, accountable for its customers' deployments, with what they learn going back into the product.