This guide is for CTOs, operations leaders and finance owners who keep hearing "forward deployed engineering" (FDE: engineers who write production code inside the customer's own team and environment) and want to know what it is, who does it and whether it fits their problem.
The term is everywhere in 2026 and people use it loosely. Vendors, consultancies and job boards apply it to different jobs. Getting a clear definition before you buy or hire can save you months, and it's the first filter against renamed staff augmentation.
What is forward deployed engineering?
Forward deployed engineering is a delivery model where a senior software engineer works inside the customer's team and environment, writes production code there, and stays accountable for a working result instead of a report. "Forward" means the engineer sits at the customer's edge, close to real data, real users and real constraints.
Four things mark it out. The engineer writes code, and that code runs in your environment, against your data and permissions. Success is judged by something running in production. And there's usually a baseline metric, agreed at the start, to measure it against.
The model exists because the hard part of most software, and especially AI software, is the last mile: integrations, edge cases, access controls, failure handling and the people who have to use the thing. A demo skips all of it. Production can't. We look at the same gap from the buyer side in why enterprise AI fails on Monday morning.
Where did forward deployed engineering come from?
Palantir popularized the forward deployed engineer role in the early 2010s, according to a16z, which describes Palantir's "Deltas" working at customer sites. For about a decade the model stayed mostly inside Palantir. It spread once AI made the gap between demo and production impossible to ignore.
That gap is well documented, with caveats. The MIT NANDA "GenAI Divide" study, reported by Fortune in 2025, found that about 95% of enterprise gen-AI pilots showed no measurable P&L impact. Critics call its methodology opaque, so take the exact number with a grain of salt. The same study found that bought tools or partnerships succeeded about 67% of the time, and internal builds about a third as often.
Gartner points the same way. It predicts that over 40% of agentic AI projects will be canceled by the end of 2027, citing cost, unclear value or weak risk controls. Putting an engineer next to the problem is one response to all three.
What are the two kinds of FDE role?
CIO.com identifies two different jobs under one title: the "product builder," who decides what to build and writes it (the Palantir model), and the "platform operator," closer to solutions engineering plus customer success, covering roughly 5-10 accounts each. Both are called FDEs, and you shouldn't treat them as interchangeable.
The split changes what you actually get. The reporting is in CIO.com's piece, there are two completely different roles called FDE.
- A product builder is embedded for weeks or months, writes the production code, makes the design decisions and owns the delivered system.
- A platform operator is spread across several customers. They configure a vendor product, help the customer adopt it, and escalate gaps back to the vendor.
Bloomberry analyzed about 1,000 FDE job postings and found the builder type in about 60% of roles, so it's the majority but far from the whole market (Bloomberry). Kaprin practices the builder model. When any provider says "FDE," ask which of the two they mean, and how many accounts each engineer carries at once.
Who uses forward deployed engineers today?
Palantir started the model. Since 2025, AI labs, hyperscalers, fintechs and consultancies have built FDE teams. OpenAI, Ramp, Microsoft, AWS, Anthropic and Deloitte are among the organizations investing, each with its own definition of the role. Scale ranges from about 15 engineers at Ramp to billions of dollars of committed funding at the hyperscalers.
The Pragmatic Engineer reports that OpenAI's FDE team began in 2025 with 2 people and now has 10 or more, and that Ramp has about 15 FDEs organized in pods embedded in product engineering teams. It also notes that each company defines the role differently.
The money is bigger at the hyperscalers. Forrester counts about $9B committed to FDE-style units by four providers between 4 May and 15 July 2026: OpenAI $4B or more, Microsoft $2.5B, AWS $1B and Anthropic's services venture about $1.5B (Forrester's own figure). CIO.com reports that Microsoft and AWS are putting billions into thousands of FDEs, and quotes analysts warning about vendor lock-in. We unpack that bet in the $9 billion FDE bet.
Consultancies are in too. Deloitte describes FDE pods of about 4-5 people, time-boxed to roughly 3 days to a use case, 3 weeks to a prototype and 3 months to production.
How is an FDE different from a solutions engineer, consultant or contractor?
An FDE writes and owns production code in your environment. A solutions engineer usually supports a vendor's sale. A consultant advises and designs. A contractor fills a seat under your direction. What differs is who writes the code, who answers for the outcome, and how the work ends.
| Role | Writes production code | Works where | Accountable for | How it typically ends |
|---|---|---|---|---|
| Forward deployed engineer (builder) | Yes, as the main output | Inside your team and cloud | A working system and an agreed metric | Handoff to your team |
| Solutions engineer | Sometimes, mostly proofs of concept | Vendor side, with customer calls | Technical win for the vendor's product | Deal closes, or moves to customer success |
| Management consultant | Rarely | Mixed, often off-site | Advice, roadmap, governance | Final report or a follow-on phase |
| Contractor or staff augmentation | Yes, under your direction | Your team, your management | Hours delivered | Contract ends, knowledge leaves |
The last two rows are where "FDE washing" happens. A provider renames contractors as forward deployed engineers without changing who directs the work or who is accountable. For a deeper side-by-side, including vendor-supplied FDEs, read FDE vs consulting vs staff augmentation.
Why is demand for forward deployed engineers rising now?
Because AI projects stall between demo and production, and buyers want engineers who can close that gap inside their own systems. Job-posting data shows a steep rise in 2026, though the exact counts differ by source, date range and how each firm defines an FDE posting.
Two reports cite Lightcast data. Fortune reports FDE job postings up more than 1,000% in January to August 2026 compared with the same period of 2025, with a median advertised salary above $188K against about $145K for software engineers. TechRadar reports about 200 US FDE postings in 2023, about 1,200 in 2025 and about 7,300 in the first nine months of 2026, and notes that three of the top four advertisers are consultancies: Deloitte, Accenture and KPMG.
The two reports use different windows, so don't try to reconcile them to the decimal. The direction is the same in both. The consultancy detail deserves a second look: some of that demand is an existing consulting or staffing role with a new label on it. That's our reading, and it's opinion, not data. The label is spreading faster than the definition.
The salary premium has a practical consequence. Hiring your own FDE-grade engineer in this market is slow and expensive, which is why many companies look for a partner who already has them.
When does forward deployed engineering fit, and when does it not?
It fits when a specific system has to work inside your data, permissions and workflows, and what you lack is engineering capacity or production experience. It doesn't fit when nobody owns the problem, when standard software already solves it, or when what you need is strategy and nobody has to build anything yet.
It tends to fit when:
- An AI pilot works in a demo but has no owner, no monitoring and no place in the daily process.
- A vibe-coded prototype is in front of real users and needs production hardening.
- A workflow touches several internal systems (ERP, CRM, payments, freight) and off-the-shelf tools stop at the edges. Our Slender Morris case moved PDF purchase orders into Cin7 and Machship, with about 90% faster order entry.
- You need a result in weeks, and you want your own team to inherit it.
It tends not to fit when:
- The problem is undefined. An engineer can't fix "do something with AI." This is the unclear-value failure Gartner describes.
- A configurable SaaS product already does the job at a fair price.
- The decision you need is about operating model or strategy, and nobody has to write software yet.
- The engineer can't get access to systems, data or the people who use them.
- You need maintenance for years. That's a permanent hire or a support contract.
What does a forward deployed engineering engagement look like?
A well-run engagement is short, fixed in scope and built to end. At Kaprin it has four stages: Scope within 24 hours, Embed on day 1, Harden in weeks 2-6, and Handoff by day 45. Production deployment typically lands within 30-45 days of starting.
- Scope, within 24 hours. We agree the problem, the baseline metric and what "done" means. Fixed-scope sprints run 4-6 weeks.
- On day 1 a senior engineer, usually in a small pod, joins your channels, repositories and standups, with access to real systems.
- Harden, in weeks 2-6. The work moves from "it runs" to "it holds up": idempotency, locks, schema-validated outputs, spend caps, audit logs, evals. Weekly live demos keep the sponsor close to progress.
- By day 45, handoff. Your team inherits the code, the runbooks and the knowledge.
You own 100% of the code, database and IP, deployed in your own cloud account (AWS or GCP) or private VPC, with SOC2-ready controls, row-level security and audit logging. The full sequence is on our how we deploy page, and the sprint format is described under forward deployed sprints.
Forrester's argument is that FDEs are "training wheels" for AI reinvention: temporary by design, with the provider's footprint shrinking as your team takes over. We agree, and it's why handoff is a named stage in our process. See handoff by design.
How do you evaluate a forward deployed engineering partner?
Check who writes the code, whether named engineers are committed, who owns the result, how the engagement ends, and what metric you agreed before work started. A credible partner answers all five in writing without hedging. Anything vague now will be vaguer once the contract is signed.
- Which of the two FDE roles do you mean, and is the engineer writing production code?
- Which named engineers will be on my team, and what happens if one leaves?
- Will it run in my cloud account, and do I own 100% of the code, database and IP?
- What baseline metric will we agree before work starts, and how do we measure it?
- What's the handoff plan, and what does week 6 look like if I never call you again?
- Can you show a production system with a named reference, beyond pilots?
- What do you do when the scope turns out to be wrong? Fixed scope should still allow an honest "stop."
A longer, buyer-focused list is in how to scope, buy and evaluate FDE services. If you'd like to test your own problem against this checklist, talk to our team. We'll tell you plainly whether an embedded engineer is the right fit or something simpler will do.
