FDE as a service, explained: The four ways AI vendors get embedded engineers to their customers

Key takeaways

• FDE as a service is an arrangement where an outside party supplies forward deployed engineers who write production code inside your customers' environments after the contract signs. The phrase covers four different arrangements: an in-house team, a partner-led program, an FDE agency, and senior engineers rented under your own badge.
• Four questions tell the routes apart: who employs the engineer, whose name faces your customer, who directs the work, and who keeps the reusable assets and what the field teaches.
• The category is young and the pricing is mostly hidden. AWS, Deloitte, Plank and CHI Software publish no price; only fde.agency and A.Team Deployed list one.
• The market's own numbers favour hiring directly: Christian & Timbers found 70% of companies planning FDE hires by the end of Q2 2026, and TechCrunch reports that companies are increasingly hiring FDEs rather than outsourcing them.
• Put the same five questions to every vendor in writing. The answers sort a partner program from an agency from a rented engineer faster than any website does.

Why this question matters

If you're an AI software company, enterprise deals are closing faster than your team can deploy them, and every supplier you talk to calls its offer FDE as a service. They aren't selling the same thing. One is a systems integrator that will carry your customer relationship, one is an agency with its own method and its own brand, one is your own hiring plan, and one is a single senior engineer who works inside your team under your direction. The route you pick decides who your customer meets and what you learn from the work.

The frame: Four questions that sort every FDE-as-a-service offer

Whatever a supplier calls its offer, four answers place it.

1. Who employs the engineer? You, a partner, an agency, or a network the engineer belongs to.
2. Whose name faces your customer? Yours, or the supplier's.
3. Who directs the work and owns the outcome? The vendor's engineering lead, or the supplier's delivery lead.
4. Who keeps the reusable assets and the learning? The code, the playbooks and the failure modes the engineer finds in the field either land in your repo and your product team, or they stay with the supplier and travel to its next engagement.

Those four answers produce four routes. They aren't ranked. Each is the right answer for some vendor at some stage, and F-Prime's point that heavy FDE effort on more than 30 to 40% of deployments is a product problem applies to all four, which is why the decision by signal lives in when to hire forward deployed engineers and this guide stays on the map.

What is FDE as a service?

FDE as a service is an arrangement in which an outside party supplies forward deployed engineers, meaning software engineers who work inside a customer's environment after the contract signs, write production code there and stay through adoption. The buyer is usually an AI software company deploying its product at its own customers, though enterprises buy it too. Plank and CHI Software both use the phrase "FDE as a Service" for their offers; AWS and Deloitte describe versions of the same thing without it.

What the label leaves open is everything the four questions ask. For the role itself, and the test for whether a given engineer is a real one, read what a forward deployed engineer is.

What is an in-house FDE team, and who runs one?

An in-house FDE team is the baseline every other route is measured against: you hire the engineers as employees, usually into engineering or product, and you keep the work, the badge and the learning. It's FDE as a service only in the sense that the supplier is the labor market.

Several of the best-documented FDE functions are built this way. Ramp, the fintech, went from two FDEs in fall 2023 to about 16, seven of them former founders. OpenAI's FDE organisation started with two people and was around 140 by August 2026, according to Gregor Ojstersek's reporting. Christian & Timbers describes the pattern at the other end of the market: pilot teams of two or three growing into organisations of 20 to 100 or more. TechCrunch's reporting on the same study adds that companies are increasingly choosing to hire FDEs directly to keep proprietary processes inside the company.

The cost is time and fixed spend. Hiring a permanent senior FDE takes six to nine months, and Perspective AI's stage table puts all-in year-one cost at $200K to $250K per engineer at seed and $450K to $550K or more at Series C. That is a different thing from a posted salary, and the build, rent or partner guide works through the dependencies.

What is partner-led FDE?

Partner-led FDE is a program in which a consultancy or systems integrator employs the engineers, holds the customer relationship and deploys under its own brand, usually with a platform vendor's credentialing behind it. The customer or the platform ecosystem buys; the vendor whose product is being deployed is often one party among several.

AWS made the structure explicit on 30 June 2026. Its partner-led FDE announcement commits $1 billion to the broader initiative, has partners build ring-fenced, AWS-credentialed engineering teams, and states that partners already hold the customer relationships and own a reusable delivery harness of ontologies, evaluation frameworks, MCP servers and a context graph. The business outcome belongs to the customer, and the compounding advantage stays with the partner. The program launches with a limited set of partners.

Deloitte announced its Forward Deployed Engineering offer in December 2025: specialists embedded with client teams in small cross-disciplinary pods, framed around a high-impact business problem with measurable success criteria. On 12 August 2026 it became the first global systems integrator to staff and deliver a Salesforce FDE engagement, with Salesforce supplying platform insight and Deloitte supplying the engineers. Neither Deloitte nor AWS discloses pricing.

The model labs have built their own versions. Ode with Anthropic is a $1.5 billion venture launched in May, built on the acquisition of Fractional AI, with about 100 engineers selling to enterprises; TechCrunch names OpenAI's Deployment Company as its competitor. TSIA's maturity model puts this kind of ecosystem scaling at its Orchestration stage, after a vendor has first learned what works with its own engineers.

What is an FDE agency?

An FDE agency is a firm that employs its own forward deployed engineers, trains them in its own method, and sells them to a vendor under the agency's brand, usually as pods and usually against a scope or a monthly term. The agency's method and its people are the product.

Plank titles its site "FDE as a Service" and sells three things: AI-native product engineering, forward-deployed AI engineering, and on-demand QA pods. Its FDE page says an embedded engineer reads your code, joins your standups and starts after a scoping call, that every engineer completes the Plank AI Accelerator, that pods can scale up as deployments grow, and that NDAs and IP assignment are in place before any code. It lists 50 or more AI companies as customers. Pricing isn't published, and the page says nothing about whose badge the engineer wears.

CHI Software is a software services firm that sells the offer. It employs the engineers, who work under the CHI brand, and sells either a fixed-scope engagement with a defined cost and timeline or a dedicated engineer billed monthly with a one-month minimum. It cites ISO 9001 and ISO 27001, says work is documented in the client's repositories, and says its own approach avoids the lock-in it attributes to other FDE-as-a-service providers. Pricing isn't published here either.

Both pages describe what the agency does. Neither tells you who directs the work or what happens to the playbooks when the engagement ends, and those are the two questions to put to any agency in writing.

What is a rented FDE under your badge?

A rented FDE under your badge is a senior engineer supplied from a vetted network who works as part of your team: introduced to your customer as your person, directed by your engineering lead, paid for at a monthly or hourly rate per engineer. The supplier doesn't appear in front of your customer, and the code lands in your repository.

This is the route the other three leave unnamed. The client directs the work and owns the outcome. The engineer owns their delivery, in the way a senior contractor does, and nobody sits above them at the supplier. It resembles the self-managed model in individual contractors vs managed teams, applied to deployment work.

A.Team Deployed sits here. A.Team puts 1 to 3 named and priced senior engineers in front of you to interview in 3 to 5 business days, at about $25K a month for a senior builder, working under your badge and your direction. fde.agency also supplies individual engineers and lists its rates on its site: $150 to $180 an hour for contract roles and $22K to $28K for fixed projects.

The route has two limits worth knowing before you choose it. Nobody at the supplier absorbs what the engineer learns in the field, so someone on your product team has to own that, or the third customer gets another bespoke build. And the route only works if the engineer is senior enough to work unsupervised in front of a customer, which is a question the FDE readiness rubric scores and the vendor's title can't answer.

zoom_in

How do the four routes compare?

The table shows each route as its suppliers describe themselves. "Not stated" means the supplier's public page doesn't say, which is itself a finding.

How do you tell which route a vendor is selling?

Ask five questions in writing before the first call, and read the answers side by side across suppliers. Who is the engineer's employer, and who pays their salary? Whose name is on the badge, the email address and the pull requests your customer sees? Who sets priorities week to week, and who signs off that a piece of work is done? When the engagement ends, what stays in your repo and what leaves with the engineer or the firm? What is the billing unit, and is there a management layer priced above the engineer?

The fifth question is the one that separates a rented engineer from a managed offer wearing the same label. If a supplier quotes delivery oversight on top of the engineer, you're buying the managed model, and you should ask where that person sits and what they cost. A partner-led program and many agencies are priced that way by construction. The layer is what you're paying for in those routes, and it should appear as a line item rather than inside a blended rate.

The websites answer some of the first four questions and none of the fifth. The pages for AWS, Deloitte, Plank and CHI Software publish no price, so the billing unit has to be asked for.

Where does each route break?

Each route fails in one predictable place, and knowing it tells you what to put around the engineer.

In-house breaks on time. Six to nine months per senior seat means the deployment backlog grows through the hiring window, and a fixed cost carries the quiet quarters. It's the right answer when deployment is the product, and Ramp's function exists because enterprise integration became its business.

Partner-led breaks on learning and badge. The partner's engineers learn your product's failure modes and carry them to the next engagement, and your customer sees the partner's name. For a vendor whose customers already expect a systems integrator in the room, that's an acceptable trade, and AWS built the program around it.

An agency breaks on method and lock-in. The agency's pods, its harness and its training are the product, and they're valuable when they match your stack and costly when they don't. CHI Software's own page names lock-in as the risk to avoid, which is the right question to ask any agency, including CHI.

Renting breaks on ownership of the loop. The engineer builds and rolls off. What gets generalised into your product depends on a named owner on your side, and what the engineer does in front of your customer depends on how senior they are.

What does FDE as a service cost?

Pricing is the least published fact in this market, and the public numbers describe different things. Perspective AI's stage table is the in-house reference: all-in year-one cost of $200K to $250K per engineer at seed, $275K to $350K at Series A, $350K to $450K at Series B and $450K to $550K or more at Series C, before recruiting. Partner-led programs and the two agencies don't publish rates. A.Team's rented route is about $25K a month for a senior builder, and fde.agency lists $150 to $180 an hour for contract roles.

Compare them on the same basis. An in-house year-one number includes equity and ramp that a monthly rate doesn't, and a fixed-scope price includes a delivery lead that a per-engineer rate leaves to you. The cost model in when to hire forward deployed engineers lays out the dependencies.

What to do next

Shortlist the two or three suppliers closest to your situation, one from each route that fits, and send them the five questions above in writing. Read the answers in a single table before you take any call. If the answers say your deployment backlog is bursty and your own team can direct the work, renting is the route to price. If they say your customers expect an integrator, price the partner. If you can't yet say which, the decision by signal is in when to hire forward deployed engineers.

When the answer is rent, A.Team puts 1 to 3 named and priced senior engineers in front of you to interview in 3 to 5 business days, at about $25K a month for a senior builder, working under your badge and your team's direction.