How to build a forward deployed engineering team inside an AI product company
Build an FDE team when enterprise customers sign but stall before going live. Start with two or three engineers who can write production code and talk to a customer's operations team, put them close to the product team so what they learn changes the product, and price their work around outcomes, not hours. Partner when you need capacity faster than you can hire.
Signals you need one now
- Deals close, then sit in implementation for months.
- Your solutions engineers are writing production code they were never meant to own.
- The same integration gets rebuilt at every customer.
- Churn risk shows up before the customer has ever gone live.
Where the team should sit
There are three common structures. A GoodFirms survey of 133 AI and software companies in mid-2026 found how often each is used:
| Structure | Share of companies | Works well when | Risk |
|---|---|---|---|
| Inside solutions engineering | 42.9% | You are early, and pre-sale and post-sale overlap | FDEs get pulled back into demos |
| Spread across business units | 32.1% | You sell into very different industries | Learning does not reach the product |
| A standalone team | 14.3% | Deployment is a major line of business | It drifts away from the product roadmap |
Whatever you pick, give the FDE lead a standing slot in product planning. The value of the role is that what breaks at one customer becomes a product change, not a custom patch. Palantir's split between Devs, who build the platform, and Deltas, who deploy it at customers, works because the two talk constantly.
Who to hire
In the same survey, companies building FDE functions most wanted software engineering skill (89.3%), product sense (85.7%), AI fluency (82.1%) and business fluency (64.3%). In practice, screen for:
- Production engineering. They ship code that stays running after they leave.
- Integration with messy systems. Real customer data never looks like the demo.
- Writing for non-engineers. They explain trade-offs to an operations team.
- Saying no. One customer's wish list can swallow the product.
Our skills table, with an interview test for each skill, is on the forward deployed engineer page.
How to price FDE work to your customers
| Pricing model | Share of companies (GoodFirms, 2026) |
|---|---|
| Hybrid | 28.6% |
| Enterprise contract pricing | 25.0% |
| Outcome-based | 21.4% |
| Retainer tied to OKRs | 10.7% |
| Fixed scope | 7.1% |
| Time and materials | 3.6% |
The pattern is clear: customers buy a result, not hours. Tie the price to the outcome you agreed before the work started.
How to measure it
The same survey found the most common success measures were pre-agreed business outcomes (78.6%), time to production (67.9%), product adoption (60.7%) and client satisfaction (60.7%). Pick two: one for the customer (an outcome) and one for you (time to production).
When to partner instead
- You need capacity this quarter and hiring will take two.
- Demand is uneven: three deployments at once, then none.
- Customers are in time zones or regions your team does not cover.
A partner's engineers can work under your direction and your process. For AMT's model, see forward deployed engineering.
FAQ
How many forward deployed engineers does an AI startup need?
Should forward deployed engineers report to sales or engineering?
What is the difference between a forward deployed engineer and a solutions engineer?
How do you stop FDE work becoming custom development?
Sources
Need FDE capacity before you can hire it? Tell us how many deployments are waiting.