Table of contents
A forward deployed engineer is a software engineer who embeds with one customer, learns how that customer works, and builds their employer’s product into the customer’s real environment until it runs in production. Palantir invented the role. OpenAI, Anthropic, Databricks and now AWS have copied it. Andreessen Horowitz has called it the hottest job in startups.
Most explainers on the role are written for engineers deciding whether to take the job, or for founders deciding whether to build the team. This one is for marketers, because the forward deployed model is about to arrive in your team in two ways. One is obvious: a vendor’s engineer turns up to deploy an AI platform into your stack. The other is quieter and, I think, more important. Marketers are becoming forward deployed themselves, taking on the engineering, design and production work that used to need a specialist, without waiting for one to be assigned.
What is a forward deployed engineer?
A forward deployed engineer (FDE) is a software engineer employed by a vendor who embeds with a single customer, learns that customer’s environment and workflows, and builds the vendor’s product into the customer’s real systems until it runs in production. They sit inside the customer’s organisation rather than the vendor’s product team, and their job is to get the product working in that customer’s real systems, with that customer’s messy data, for that customer’s actual workflows. They scope the problem, write production code, run evaluations, roll it out, and feed what they learn back to the people building the core product.
The best one-line summary comes from Palantir itself, which called its FDEs “Deltas” and its product engineers “Devs”:
“You can think of a Dev’s focus as ‘one capability, many customers,’ while a Delta’s focus is ‘one customer, many capabilities’.”
Palantir, Dev versus Delta: Demystifying engineering roles at Palantir, as quoted by The Pragmatic Engineer
That distinction is the whole role. A product engineer builds one feature and ships it to everyone. An FDE builds whatever one customer needs, in that customer’s environment, and owns it until it works.
Where the role came from
Palantir created the role in the early 2010s out of necessity rather than org design. Its early customers were intelligence and defence agencies with data environments so sensitive and idiosyncratic that remote delivery was impossible. Engineers had to sit on site, often on air-gapped networks, learn the domain, and build there.
According to The Pragmatic Engineer, until around 2016 Palantir had more FDEs than it had conventional software engineers. When the company launched its Foundry platform that year, many of those FDEs moved into core product work and brought years of field knowledge with them. That feedback loop, field to product, is the part most copycats miss.
What an FDE does day to day
The day-to-day breaks into five stages. If you run campaigns as plan, launch, analyse, repeat, the shape will look familiar:
- Discovery. Sit with the customer, learn how they define their nouns and verbs, find the real problem behind the stated one.
- Technical scoping. Work out what the product can do in this environment and what needs to be built around it.
- Build. Write production code inside the customer’s systems, not a demo on a laptop.
- Evaluate and roll out. Test against real data, measure, fix, ship to real users.
- Feed back. Report what the platform could not do to the product team so the next deployment starts further ahead.
Discovery and scoping are the plan stage: understand the environment before you commit to building anything. Build, evaluate and roll out are launch: put it into the real environment with real data and real users, and fix what breaks. Feed back is analyse: work out what the platform could not do and pass that on. And then repeat, with the next deployment starting from what the last one learned.
That last stage is the one marketing teams most often skip. Campaigns get planned and launched, but what they revealed rarely makes it back into the next plan. FDEs treat feeding back as the reason the job exists, which is why the model compounds.
Colin Jarvis, who leads forward deployed engineering at OpenAI, described why the build stage is so different from ordinary product work:
“FDEs work in a ton of ambiguity, and often what the customer describes in scoping doesn’t match the data/system reality on the ground.”
Colin Jarvis, OpenAI, in The Pragmatic Engineer
Any marketer who has been handed a “clean” CRM export will recognise that sentence.
Why every AI company suddenly has FDEs
The role stayed a Palantir curiosity for a decade. Then large language models made it the default enterprise AI delivery model, because the hard part of enterprise AI turned out to be deployment, not the model. Requirements are vague, data is messy, and evaluation only happens in production. That is precisely the environment the FDE was designed for.
The scale of the shift is now visible in the hiring numbers and the balance sheets:
- AWS announced a $1 billion forward deployed engineering organisation in June 2026, embedding thousands of engineers with customers to build agentic AI systems. Francessca Vasquez, AWS VP of Frontier AI Engineering and Services, framed the goal as “Customers leave AWS FDE deployments with both new solutions and new engineering capabilities.”
- OpenAI, Anthropic, Databricks, Cohere and Scale AI all run FDE teams. Ramp, a fintech scaleup, built a team of around 15 FDEs in under a year, according to The Pragmatic Engineer.
- Palantir alumni seed most of them. Roughly 250 engineers went through Palantir’s Project Frontline rotation, and many now lead FDE teams at OpenAI, Anthropic, xAI and Anduril, according to Vinoo Ganesh, a former Palantir engineer and now CEO of Kepler.
Ganesh’s own definition is the most useful one for what follows, because it names the thing the role is really for:
“An FDE solves customer problems in order to earn the insight that informs what gets built next.”
Vinoo Ganesh, CEO and Co-Founder, Kepler
Hold on to that word, insight. The code is the price of admission. The insight is the product.
Forward deployed engineer vs GTM engineer vs growth engineer
Marketing already has a crowded shelf of engineer-adjacent titles, and “forward deployed” is about to be added to it. The distinctions matter because they change who pays, who they report to, and what they leave behind.
| Title | Employed by | Primary focus | Sits closest to | Leaves behind |
|---|---|---|---|---|
| Forward deployed engineer | The vendor | Deploying one vendor’s product into one customer’s environment | The customer’s operations team | A working deployment of that vendor’s product |
| GTM engineer | Your company | Automated revenue systems: CRM, enrichment, outbound sequencing | Sales and RevOps | Vendor-neutral workflows you own |
| Growth engineer | Your company | Product-embedded growth mechanisms: referrals, onboarding, activation | Product | Features and loops in your product |
| Marketing engineer | Your company | Tracking, attribution, landing pages, martech plumbing | Marketing | Infrastructure behind campaigns |
The distinction that carries most of the weight is FDE versus GTM engineer. The FDE works for the vendor and builds that vendor’s product in, while the GTM engineer works for you and builds across whatever stack you have. If a vendor offers you an FDE, you are getting a very good engineer whose incentives point at their platform, not yours. That is fine as long as you know it.
The forward deployed marketer
The title has started to spread beyond engineering, exactly as a16z predicted it would. Three distinct jobs now share it.
The vendor’s customer marketer. Cognition, the company behind the Devin coding agent, is hiring a forward deployed marketer to embed with a portfolio of major accounts, find the stories of business impact, and turn them into press, brand films and written case studies. Windsurf has advertised the same title. This is customer marketing done from inside the deployment rather than from a briefing document.
The agency’s embedded builder. Eric Siu’s Single Grain hires forward deployed marketers who work inside a client’s stack, build AI agents for research, content operations and reporting, and are measured on the client’s business KPI. 42 Agency calls its version “forward deployed GTM”, and Myosin describes “forward deployed marketing” as build, install and transfer. All three are consultancies that have noticed clients want running systems, not decks.
The internal enabler. This is the one that matters most for the argument in this piece. Stripe is hiring Forward Deployed AI Accelerators for its marketing team, and the job description is worth reading closely because it is not about deploying a product into a customer. It is about deploying the forward deployed model into Stripe’s own marketers:
“We’re building a team of Forward Deployed AI Accelerators (FDAs) embedded directly with marketing cohorts to make AI the default mode for all work — not an occasional tool, but the foundation of how every marketer at Stripe executes.”
Each accelerator is embedded with a cohort of 25 to 45 marketers. The responsibilities read like an FDE job description with the customer swapped for a colleague: “Build custom tools, agents, automations, and skills tailored to each marketer’s specific responsibilities”, then “Teach marketers to build and iterate on their own tools over time, creating independence”. The stated end state for every marketer in the cohort is “self-sufficiency”. The requirements list “Proficiency with AI development tools and platforms (e.g., Claude Code, Codex, custom agent frameworks, API integrations, workflow automation tools)”.
Read that again. A payments company is paying people to sit with its marketers until those marketers can build their own tools and no longer need them. That is the forward deployed model applied to marketing, and the deliverable is a marketer who does the engineering.
Some of the titles above are arbitrage. Tom Hollands at a16z offers a fair test: does the title describe work that would have been unrecognisable five years ago? A product marketer who sits in on customer onboarding calls is not a new job. A marketer who then builds the onboarding email sequence, the in-app checklist, the usage dashboard and the demo video, all in the same week and without filing a ticket, probably is. Stripe’s cohort is being trained to be the second kind.
The deployment is running the other way
The forward deployed model exists because two things are scarce at the same time: deep domain context and the ability to build. Palantir’s insight was that you cannot ship domain context remotely, so you post the engineer into the domain and let them absorb it.
Marketing has the opposite problem. Marketers already have the domain context. They know the customer, the message, the funnel, the objection on the third sales call. What they have never had is the ability to build. Every landing page, tracking change, calculator, video or design asset went through a queue owned by someone else. The marketer was forward deployed into the domain from day one. The missing half was the engineering.
That half has now collapsed. In the teams I work with, marketers are taking on a steadily larger share of the engineering role themselves, and vendors are embedding far fewer engineers than the job adverts suggest:
- Pushing code changes to git. Marketing sites live in repositories now, and marketers are editing them directly with AI coding tools rather than briefing a developer. This article was written in a code editor and pushed to a git repository by a marketer. No engineer touched it.
- Building small apps as engineering as marketing. Calculators, graders, generators and diagnostic tools that used to be a quarter of engineering time are now a weekend of vibe coding. The satellite app pattern, one small tool per funnel stage, is only possible because the marketer can build it.
- Taking on design. Layouts, social assets, diagrams and ad creative that used to wait on a designer are produced in the flow of the campaign, by the person who understands the message, and refined by a designer only where it earns the time.
- Producing video. Screen recordings, product walkthroughs, talking-head explainers and cut-downs, produced and edited without a production team, because the tooling no longer requires one.
- Wiring up their own data. Connecting analytics, CRM and ad platforms through MCP servers and agents rather than waiting for a data engineer to build the pipeline.
None of this needs additional expertise in the sense of a new specialist on the payroll. It needs the marketer to work the way an FDE works: scope the problem, build in the real environment, evaluate against real data, ship, and feed back what you learned into the next campaign.
This is the practical end of the engineeringification of marketing. That piece argued marketers need to adopt the engineering mindset. The forward deployed engineer is what the mindset looks like as a job description, and marketers are quietly taking it on without the title.
The insight loop still matters
Ganesh’s warning for FDEs applies word for word to marketers who build their own things:
“Do the work with nothing underneath it and you learn one company’s model, ship something shaped exactly to it, and lose all of it when the engagement closes. The next customer starts from zero, and so does the one after that. That’s consulting.”
Vinoo Ganesh, CEO and Co-Founder, Kepler
Swap “customer” for “campaign” and you have the trap most marketing teams fall into with AI tooling. Every campaign is built from scratch, every asset is a one-off, and nothing compounds. The value of building your own tools sits in the reusable system, skill or template that means the next campaign starts further ahead. A calculator is a one-off. A calculator pattern you can reproduce for any segment in an afternoon is an asset.
That is also why the vendor FDE model deserves some caution from marketing leaders. Thomas Otter, an enterprise software investor, put the financial version of the problem bluntly:
“Support isn’t R&D, no matter how technical it gets.”
Thomas Otter, WTF is a forward-deployed engineer?
His argument is that FDE work is a services cost dressed as product, and a company that leans on it too heavily struggles to build standard product later. The marketing version is the same. A marketer who builds bespoke things for every campaign and never turns them into systems is running a services function inside marketing. A marketer who builds once and reuses is running an engineering one.
What to do about it
If the forward deployed model is coming to your team either way, here is how I would prepare for both directions.
When a vendor offers you an FDE:
- Ask who they report to. Ganesh’s rule is that FDEs should report to product, not sales, so that every deployment produces something the next one starts from. A vendor FDE who reports to the account team is there to close the account.
- Decide up front what stays when they leave. Insist the working system, the documentation and the skills transfer to your team. AWS made “leaving customers self-sufficient” a stated goal of its programme. Hold every vendor to the same standard.
- Keep your data and workflows portable. An FDE will build on their platform by definition. Make sure the tracking, the content and the campaign logic are yours and would survive a change of vendor. Staying model and vendor agnostic is a lot cheaper to decide before the engineer arrives than after.
When your marketers are becoming the FDE:
- Give them the access. Repository access, analytics access, a sandbox to deploy to. Most of the engineering a marketer takes on is blocked by permissions, not skill.
- Make building a normal part of the campaign, not a special project. Scope, build, evaluate, ship and feed back is just plan, launch, analyse with a build step in the middle.
- Treat every one-off as a candidate for a system. When a marketer builds a tool, a template, an agent skill or a video workflow, ask what the reusable version looks like and capture it. That is the insight loop, and it is what turns forward deployed marketers into a compounding advantage rather than a busy one.
- Stop waiting for the title. Nobody needs to be called a forward deployed marketer to work like one. The marketers doing this today are called content leads, growth managers and marketing directors.
About Growth Method
Growth Method is built for the marketer who has become the builder. It is the agentic marketing platform for B2B teams, with pre-built AI agents that plan, launch and analyse campaigns from your live marketing data, so the scoping, building and evaluating an FDE does for one customer happens inside your own campaign workflow. It connects to your analytics, CRM and site through MCP, is model and vendor agnostic, and captures what each campaign teaches as reusable skills the next one starts from.
We are on-track to deliver a 43% increase in inbound leads this year. There is no doubt the adoption of Growth Method is the primary driver behind these results.
Get started to connect your stack and run your first agent-planned campaign.
Frequently asked questions
What is a forward deployed engineer?
A forward deployed engineer (FDE) is a software engineer who embeds with a single customer, learns their environment and workflows, and builds their employer’s product into that customer’s real systems until it runs in production. Palantir created the role, calling these engineers Deltas, and summarised the difference as one customer, many capabilities for an FDE versus one capability, many customers for a product engineer.
How is a forward deployed engineer different from a GTM engineer?
A forward deployed engineer is employed by a vendor and posted into a customer to deploy that vendor’s product. A GTM engineer is an internal hire who builds automated revenue systems across the company’s own CRM, enrichment and outbound stack. The FDE is temporary and vendor-specific. The GTM engineer is permanent and vendor-neutral.
What is a forward deployed marketer?
A forward deployed marketer is a marketer who works inside the real operating environment rather than from the marketing function: embedded with customers, sales or product, building systems and content from what is happening. The title is new and still loosely defined. In practice the more common version is not a new hire but an existing marketer who has taken on the engineering, design and production work that used to need a specialist.
Do marketers need to learn to code to work like a forward deployed engineer?
No. The point of the forward deployed model is domain context plus the ability to build. Marketers already have the context. AI coding tools, vibe-coding platforms and agentic marketing platforms now handle enough of the build that a marketer can push a site change to git, launch a small tool or produce a video without a specialist. The skill that matters is describing what should be built clearly and checking the result.