rokan.

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

A forward-deployed engineer takes the product live at each new customer: connecting it to the customer’s systems and data, configuring it for their workflows, testing it and staying through launch. In 1,685 public FDE, deployment and implementation postings Rokan counted in November 2026, 71% describe connecting to customer systems and data and 59% describe per-customer configuration.

Updated · Sources linked inline

What the job descriptions say

Rokan counted keywords across 1,685 public forward-deployed, deployment and implementation postings at 673 companies, as of November 4, 2026.

Share of postings that describe each kind of work. Keyword counts, so read them as direction, not precision.
Work describedPostings
Connecting to the customer’s systems and data71%
Configuring workflows per customer59%
Tuning prompts and evals per customer41%
Cleaning customer data12%
Telephony3%

The job is mostly integration. That is why go-lives stall on access, not on the model.

A typical go-live, week by week

  1. Week 0. Scope one workflow, agree what “live” means and the pass bar, and send the access requests.
  2. Week 1. Access, sandbox, the history export and field mapping. Someone chases approvals every day.
  3. Week 2. Replay against the customer’s past calls or tickets, fix, rerun. The security questionnaire and DPA run in parallel.
  4. Week 3 and on. Shadow mode, then 10%, 50% and 100%, then on call for whatever breaks. See how to roll out safely.

That shape is an estimate from the postings and vendor docs, not a measurement. Systems that need vendor approval stretch week 1 into months; see how long a go-live takes.

The work nobody is hired for

In a narrower set of 929 deployment postings, the security review appears in 6%, acceptance testing in 4%, the DPA in 4% and HIPAA or a BAA in 3%. Evals appear in 29% and monitoring in 21%. The rarely mentioned items are the ones that stall go-lives, so they usually land on the founder, at night.

What it costs

The median FDE base salary at AI startups is about $181,000 (Recruiting From Scratch). Larger companies run whole programs: Decagon says its FDEs take about 6 weeks per go-live (Let’s Data Science (news report, confirm)), Atlassian launched an FDE program with 12-week engagements (BusinessWire), and OpenAI launched a deployment company in May 2026 (OpenAI). At a seed-stage startup, the first FDE is usually the CTO.

When to hire one

Hire an FDE when go-lives are steady enough to keep one person busy and you want that knowledge in-house. Until then, the same work still has to happen for every signed customer, usually by a founder.

Short answers

What is a forward-deployed engineer?

An engineer who works with each customer to take the product live in their environment: connecting systems and data, configuring workflows, testing and supporting launch.

How much does a forward-deployed engineer earn?

The median base salary at AI startups is about $181,000 in 2026, according to Recruiting From Scratch.

What do forward-deployed engineers spend most of their time on?

Integration. In 1,685 open FDE, deployment and implementation public postings Rokan counted in November 2026, 71% describe connecting to the customer’s systems and data, 59% per-customer configuration and 3% telephony.