What is a forward deployed engineer? The job, without the hype.
A forward deployed engineer (FDE) is a senior software engineer who works inside a customer's company and builds what that one customer needs. They sit with the people doing the work, decide where AI belongs, build it into the systems the company already runs and hand it over. It is how a stalled AI pilot reaches production.
The title is everywhere because the money is. The big AI labs and cloud providers put billions behind forward deployed engineering in 2026, and US job postings for the role went from about 200 in all of 2024 to more than 5,200 between January and July 2026, according to Lightcast.
Full disclosure: I'm Christian Vismara, co-founder and CTO of Digital Kitchen, and I sell this work, through the studio and as an interim FDE in New York. So I can tell you what the job looks like on an ordinary Tuesday, and when hiring one is a waste of money.
Where did the forward deployed engineer title come from?
Palantir made the title popular. Its forward deployed engineers were embedded at the customer and customized the platform for each client, and the name comes from military language, which fits: the engineer gets posted out to where the customer's work happens.
A common way to put the difference: a product engineer builds one capability for many customers, and an FDE builds many capabilities for one customer. The FDE answers to one customer's problem first and to the product roadmap second.
It also explains the travel. OpenAI's FDE job ads mention travel of up to 50%. Palantir expects its FDEs to spend around 25% of their time on site, according to The Pragmatic Engineer. Nobody learns how a claims team really works over a video call.
What does a forward deployed engineer actually do?
An FDE turns one messy business process into software that runs inside your existing systems. A real week has less coding in it than you'd expect, and a lot more time next to the people who process the invoices or answer the tickets.
Sit with the people doing the work. The accounts payable clerk, or the ops manager who knows why one supplier's invoices always get checked twice. That knowledge is the real spec.
Map the real process next to the documented one. The documented version lives on a wiki page. In real life there is a side spreadsheet and at least one exception nobody wrote down.
Decide where AI belongs. Each step gets one of these: a model, plain code, a human approval where money or risk is involved, or the delete key. The cheapest AI feature is the step you remove.
Build inside the client's systems. The ERP, the CRM, the helpdesk, the Slack channel where approvals already happen. A new dashboard nobody opens is how pilots die.
Write evals. A set of real cases with known right answers, run on every change, so "it seems better" turns into a number you can check.
Run it in shadow mode. The system does the work next to the humans for a while, and you compare its output with theirs before anyone relies on it.
Hand it over. Docs, a runbook, the eval suite and someone on the client's team who can change the system without calling the FDE.
The sitting and the mapping are the hard part, and exactly what a pilot built on demo data skips.
Why is everyone hiring forward deployed engineers in 2026?
Selling access to a model turned out to be the easy part. Getting it to do useful work inside a real company is harder, and in 2026 the biggest names in AI started paying for that.
| Who and when | What they committed |
|---|---|
| OpenAI Deployment Company11 May 2026 | Over $4B raised from 19 investors with TPG in the lead, plus the purchase of Tomoro and its roughly 150 FDEs |
| AWS30 June 2026 | $1B for an FDE organization meant to leave customers running things on their own once a deployment wraps up |
| Microsoft "Frontier Company"2 July 2026 | $2.5B, with 6,000 experts placed inside customer companies |
| Ode (Anthropic with Blackstone and Hellman & Friedman)15 July 2026 | About $1.5B, running on Fractional AI and its roughly 100 engineers, with mid-size companies as the target |
Salesforce has promised a 1,000-person FDE team on top of that. Job postings tell the same story: Lightcast counted about 1,200 US FDE job postings in 2025 before the 2026 jump. Fortune, citing Lightcast, put the growth at about 1,000% year on year for January to August 2026, while tech postings overall grew 13%.
Supply didn't keep up. Gartner puts the number of working FDEs at only about 2,000, and a recruiter quoted in The Pragmatic Engineer says maybe 10% of engineers want the role. I can see why. The job means travel and long sessions with people who aren't engineers, and you own results that depend on someone else's messy data.
It pays well, and less of the week is code than you'd expect. The pay data is in what a forward deployed engineer costs, and the trade-off with product and consulting roles is in FDE vs consultant vs solutions engineer.
Is forward deployed engineering just consulting with a new name?
Sometimes it is, and the sharpest version of that criticism comes from Gartner.
In a 30 September 2026 report covered by The Register, Gartner predicted that by 2028, 70% of enterprises will walk away from agentic AI that vendor FDEs built for them, as the costs keep rising and the companies can't change those systems on their own. It also expects fewer than 20% of FDE engagements to make it back into the vendor's own product, and it warned about "FDE washing": plain consulting with an FDE label on it.
Sellers say much the same. The integrator SHI and the vendor Cisco told The Register in October 2026 that embedding engineers at customers is nothing new, and the same piece notes that DoIT moved its long-running customer reliability engineering practice under the FDE name in late 2025.
I agree with most of it. If the vendor's engineer leaves and nobody on your team can change the system, you bought an expensive demo. Gartner's own advice is the right checklist: agree on scope, incentives, governance and ownership on day one, along with knowledge transfer and an exit plan. I would put code ownership first, in writing, before anyone writes a line.
Does my company need a forward deployed engineer?
You need one when a live business process runs through your own systems and the AI pilot meant to fix it never reached production. If a product you can buy already does the job, you don't.
- Your pilot worked on demo data and fell apart on real data.
- The process crosses several systems and nobody owns the whole thing.
- Your engineers are busy with the product and have no time to sit with operations.
- Money or risk is involved, so the AI needs a human approval step and an audit trail.
- A product you can buy already does it. Read build vs buy AI first.
- You haven't picked a process yet. An FDE needs a target.
- You want a strategy document. That is a consultant's job (see forward deployed engineer vs consultant).
- A competitor announced agents and now you want some too. Start with what AI agents can do for a business.
How do you get a forward deployed engineer?
You can hire one, bring one in on an interim basis, or hand the work to a studio that embeds its own engineer. The right choice depends on how permanent the work is and how soon it has to run in production.
Hire one full-time
Right for permanent work close to your product, if you can wait: FDE searches take 8-12 weeks, according to the recruiting marketplace Paraform. I cover the search in how to hire a forward deployed engineer and salaries in what a forward deployed engineer costs.
Bring one in on an interim basis
Use this when you need production results this quarter. It is what I do: I join your team 2-3 days a week, on site in New York or remote, for $16,000-$20,000 per month, with the days and the first 30-day goal agreed up front. In the city? See forward deployed engineer in NYC.
Hand it to a studio
When the work needs more than one engineer, I bring Digital Kitchen, the AI studio I co-founded. A senior engineer sits inside your team, and the rest of the studio works behind them on the same workflow.
Who I am, briefly
I'm Christian Vismara, co-founder and CTO of Digital Kitchen and a software engineer in Astoria, Queens, with 11+ years of building software and an MSc in Computer Science Engineering from Politecnico di Milano. Along the way I've built 10+ AI products end to end, from WhatsApp check-ins for coaches to an AI recovery app.
At Chatterbox, a corporate language learning company, I took over as lead engineer and rebuilt the platform through a seed round and the pivot from B2C to B2B. As fractional CTO for LaunchPharm, a pharma marketing platform used by 2,000+ healthcare professionals in 67 countries, I moved the product off vendor-locked iframes to self-hosted Kubernetes. Earlier, at NTT DATA, I rebuilt Sky Italia's "Fai da Te" app, and its store rating went from 3.0 to 4.5.
Frequently asked questions
What does forward deployed engineer mean?
A forward deployed engineer is a software engineer who works inside a customer's company and builds what that customer needs. Palantir popularized the title, which borrows from military language. A product engineer builds a feature once for every customer, while a forward deployed engineer builds whatever one customer needs.
What is the forward deployed engineering model?
It is a way of delivering software where the engineer works inside the customer's company. They map how the work is really done, decide where AI belongs and where a person must approve, build it into the company's existing tools, test it with evals and shadow mode, then hand it over. It only works if the handover is planned from day one.
How much time does a forward deployed engineer spend on site?
It depends on the employer. OpenAI's FDE job ads mention travel of up to 50%. Palantir expects its FDEs to be on site around 25% of the time, according to The Pragmatic Engineer. As an interim FDE in New York, I agree the on-site days with each client up front and work remotely the rest of the time.
Is a forward deployed engineer just a consultant with a new title?
Sometimes. In September 2026 Gartner warned about "FDE washing", plain consulting sold under the FDE name, and The Register reported that DoIT moved its long-running customer reliability engineering practice under the FDE name in late 2025. A real forward deployed engineer writes and deploys code inside your systems, then hands it over. If the main deliverable is a slide deck, you are buying consulting.
Can a startup or small company get a forward deployed engineer?
Yes. Paraform reports that 59% of the companies hiring FDEs through its marketplace are seed to Series A. If a full-time hire is too slow, an interim FDE can work part of the week inside your team. I do this 2-3 days a week for $16,000-$20,000 per month, or $175-$250 per hour for scoped pieces.
Got an AI pilot that never reached production?
Book a 30-minute call and tell me where it got stuck. I'll tell you whether you need an FDE at all and, if you do, what shape the work should take. Prefer to start on your own? Take the free AI Reality Check first.