Why Everyone Is Suddenly Hiring Forward Deployed Engineers
The role at the top of the world's most expensive AI companies' job pages right now is the Forward Deployed Engineer. Look at who is hiring, and the list is unusual.
OpenAI is actively recruiting in Munich. Salesforce has FDE roles open across Munich, Jena, Düsseldorf, Berlin, Mannheim, and Frankfurt simultaneously. Anthropic, Palantir, and Cohere are hiring in parallel. Boston Consulting Group (BCG) X has a Forward Deployed AI Engineer role open in Munich. In April 2026, Accenture and Microsoft launched a joint Forward Deployed Engineering practice.
The accounting firm Ernst & Young (EY) United Kingdom (UK) and Ireland rolled out their own FDE roles that same month. According to Computerwoche, FDE postings grew 729 percent year-over-year between April 2025 and April 2026, reaching 5,330 open roles globally.
This is not Human Resources (HR) marketing. It is a quiet admission. AI sells well. It does not arrive in real work. 88 percent of enterprise AI pilots never make it to production (industry estimate, 2026). The reason is rarely the model. The reason is that nobody on site understands how the client actually works and where the AI should go.
"In the mid-market, this carries." — Jamin Mahmood-Wiebe, IJONIS Hamburg.
This post explains what a Forward Deployed Engineer actually does, where the role comes from, why it is exploding right now, and why mid-market companies need it without being able to hire one themselves.
What a Forward Deployed Engineer Actually Does
Picture someone who, in the morning, does not drive to their own office but to yours. First, the engineer sits next to your accounts team and watches 340 invoices being checked by hand each week. Then, the same person follows your sales team through a customer call and sees how the quote ends up in a spreadsheet after the meeting. An hour at the support desk is enough to notice that the same three questions need six different tools to answer.
Only then do they open the laptop and start building. A small automation at the point where one hour per day vanishes. An agent that answers the same three questions before anyone opens a ticket. A connection between the Customer Relationship Management (CRM) system and accounting, so the quote does not have to be retyped.
An FDE is three things in one person. Listener, who can read processes. Builder, who ships code. Operator, who stays until it runs in daily use. Not strategy. Not a prototype. AI that actually works inside real workflows.
Where the Term Comes From: Palantir, OpenAI, Anthropic
The role is not new. Palantir invented it more than a decade ago after discovering that analytical software for intelligence agencies and large corporations cannot be sold from the outside. Someone had to go in, see operational reality, and tailor the product to it.
OpenAI and Anthropic adapted the model for the AI wave. Large Language Models (LLMs) are generic tools. Value only appears when someone carries them into a specific accounts team, a specific logistics floor, a specific HR department. The foundation model companies realised the models alone were not enough and have hired hundreds of FDEs since.
Sequoia put a name on it in April 2026 with their thesis "Services are the new software": AI agents replace vertical software, companies pay for completed work, not for tools. If you sell finished work instead of tools, you need people who understand on the ground what the work actually is. Those people are the FDEs.
Why the Role Is Exploding Right Now
Three independent developments collide in 2026 and turn the Forward Deployed Engineer from a Palantir niche role into a category that every major foundation model lab and consulting house is racing to staff. Each one on its own would not be enough; together they make embedded delivery the only model that actually closes the gap between AI capability and real workflow impact.
One: AI tools are mature but misunderstood. Language models, agent frameworks, and automation tooling are good enough in 2026 that almost anything is technically possible. At the same time, almost no company has people who can judge which workflow is even AI-suitable. The gap between technically possible and operationally shipped has never been larger.
Two: Building software has become cheap. An internal application that cost six figures in 2024 now ships in one or two days. The expensive bottleneck is no longer the code. It is the question of what to build. That question is not answered in a meeting room. It is answered by someone who sits next to the team for two weeks.
Three: Traditional consulting cannot fill the gap. Classical consultants write strategy documents and hand them to an implementation department. The step between strategy and code is removed under the FDE model because the same person does both. Accenture, EY, and Microsoft figured that out and built their FDE practices. But they target enterprises with budgets in the hundreds of millions.
The Mid-Market Problem: Need the Job, Cannot Get the Engineer
The German Mittelstand and mid-market companies elsewhere are exactly the market that needs the role most urgently. A 200-person manufacturer, a family-owned business, a logistics company runs 40 to 60 software tools, often glued together by spreadsheets. That is precisely the reality an FDE belongs in.
But it does not happen. The compensation numbers explain why: a mid-level FDE at a foundation model company sits at $300,000 to $450,000 total comp, with senior FDEs clearing $500,000. Frankfurt investment banks and DAX corporates can match those numbers. Mid-market businesses cannot.
Three structural reasons reinforce the gap:
Vorteile
- FDE talent concentrates in San Francisco, London, and Berlin startups
- Enterprise consulting practices start at half-a-million budgets
- Mid-market companies lack both headcount budget and brand pull to hire their own FDEs
Nachteile
- At the same time the largest efficiency reserve: one hour per employee per day
- At the same time the lowest appetite for another large project
- At the same time the highest competitive pressure from cheaper international rivals
The companies that would benefit most from an FDE engagement are the hardest ones to actually staff one into.
A handful of providers have noticed the gap. Pexon Consulting offers FDE services as an Amazon Web Services (AWS) and Microsoft partner but targets banks, industry, and energy clients from 300 employees upward.
SaaStr's piece "Do FDEs work for SMBs?" leaves the question open because the classic FDE economics do not pencil out for a 200-person business. Utsubo coined "Forward Deployed Studio" for an 11-500 employee pod model.
That gap is exactly where IJONIS sits for German, Austrian, and Swiss (DACH) mid-market clients: FDE-grade delivery, agency-style packaging, mid-market sizing.
How FDE-as-a-Service Reaches the Mid-Market
The honest answer is that the job itself stays the same. Only the delivery model changes.
A mid-market business cannot match Silicon Valley salaries for an in-house FDE. A 200-person company also cannot absorb a half-a-million enterprise consulting engagement.
What works at this size is different.
A small embedded team, delivered as a service, sized to mid-market budgets, with a fixed scope and an end date. In practical terms:
- Small team, embedded on site. Two to three hands-on people, not a 20-person consulting department.
- Fixed window, clear milestones. 8 to 12 weeks from day one to a working solution, not open-ended, not delayed.
- Clean handover to the client team. At the end, the system runs without us. IJONIS stays as a sparring partner instead of a permanent dependency.
This is exactly how IJONIS has worked since 2024. Both founders are hands-on, nobody sits in the background.
Operational backbone is our term for the mid-market-shaped Company Operating System (OS). See also What Is a Company OS?. The model is built this way: discovery on site, building in two to eight weeks, handover to people who carry it forward themselves.
The Concrete Workflow at IJONIS
A Forward Deployed Engineer engagement at IJONIS follows the same four-phase shape every time, regardless of industry or technology stack. The phases are deliberately short, deliberately observable, and deliberately handed back to the client team at the end. Below is the typical timeline for a mid-market engagement of 8 to 12 weeks.
| App | Approach | Duration | Outcome |
|---|---|---|---|
| 1. On-site discovery | Work alongside, observe, ask | 3-5 days | Map of all workflows, three prioritised AI use cases |
| 2. Fast prototype | Something working for the highest-value use case | 1-2 weeks | One real improvement, used by a real team |
| 3. Scale out | Build out the other two use cases | 4-6 weeks | Three running workflows, measured time saved |
| 4. Handover | Documentation, training, mentor model | 1-2 weeks | Team owns it independently, IJONIS stays as sparring partner |
This is not a strategy document sitting in SharePoint. This is AI working inside real workflows after 8 to 12 weeks.
What an FDE Is Not
The Forward Deployed Engineer role gets confused with three established jobs that look superficially similar but operate on fundamentally different mechanics. Each comparison below clarifies the boundary so a buyer can decide which role the situation actually needs, instead of hiring the wrong shape of help and being disappointed.
An FDE is not…
…a consultant. Consultants write strategy and leave. FDEs write code and stay until it runs.
…a freelancer. Freelancers wait for a specification. FDEs invent the specification by watching.
…an implementation partner for an off-the-shelf product. Implementation partners configure what already exists. FDEs build what does not exist yet, because every company works differently.
The point is not that the other roles are worse. They are different. Rolling out standard software does not need an FDE. Putting AI into your own workflows does.
Why "Embedded" Is the Difference
Most AI projects do not fail on technology. They fail on a simple structural mechanism that has nothing to do with the model and everything to do with where decisions are made. Someone decides in a meeting room where AI should help. Someone else implements it somewhere else entirely. Nobody in between makes sure the solution actually fits the work that real people do every day.
Working embedded breaks that mechanism. Standing next to the accountant when she opens the third tab in the spreadsheet, you see the real problem. Listening during the customer call, you hear what is actually asked. Sitting next to the warehouse for a week, you notice the supplier email arrives at 6:30 and nobody touches it until 9:30.
No workshop produces this insight. No requirements analysis captures it. Only presence does. Which explains why the FDE role is being rolled out everywhere right now: presence is the only known answer to the fact that AI without context is worthless.
When FDE-as-a-Service Is the Right Fit
Not every mid-market company benefits from a Forward Deployed engagement, and pretending otherwise wastes everyone's time. Three simple checks separate companies where an embedded AI engineer will produce real value within 8 to 12 weeks from companies that need a different kind of help first.
-
Do you run at least three central software systems? CRM, Enterprise Resource Planning (ERP), accounting, warehouse management, ticketing in any combination. Fewer than three is usually too small for a meaningful engagement.
-
Is there at least one weekly question that nobody can answer in 30 minutes? "How many open quotes are outstanding right now?" "Which customer needed the most support in the last three months?" "Where do our open complaints stand today?" If questions like these cost half a day, operational backbone pays for itself.
-
Would anyone act on the answer? If the answer helps nobody because nothing happens afterwards, you do not need AI. You need clarity on roles. That is a different conversation.
Three yeses, and FDE-as-a-service is the right path.
Frequently Asked Questions About Forward Deployed Engineers
How does a Forward Deployed Engineer differ from a classical Information Technology (IT) consultant?
A consultant delivers analysis and strategy and hands implementation to someone else. An FDE is observer, architect, and developer in one. The person who understands the problem also writes the code that solves it and stays until it is in productive use.
How long does a typical FDE engagement in the mid-market take?
A full engagement at IJONIS from initial discovery to clean handover runs 8 to 12 weeks. Shorter pilot engagements of three to four weeks are possible when a single workflow is being tested in isolation. Engagements longer than three months are rare and usually signal that the original scope was cut wrong.
What does FDE-as-a-Service cost in the mid-market?
In summary, the market rate in 2026 for an 8-12 week embedded engagement sits between 30,000 and 80,000 euros. That places the model well below classical enterprise consulting projects starting at 500,000 euros and inside the typical mid-market budget for a new software initiative.
What preparation does the client company need to do?
The preparation effort on the client side is intentionally minimal. The bottom line: access to the three to five most important systems, an operational point of contact from daily business, and the willingness to let an external person watch how the work happens. A prewritten requirements specification is explicitly not needed.
What happens to the solution after the FDE engagement ends?
The key takeaway is that the client team takes over operations. IJONIS remains available as a sparring partner for adjustments, further development, and reviews, without creating a permanent dependency. The knowledge sits with the client at the end, not with the vendor.
Key Takeaways: New Role, Old Job
Forward Deployed Engineer sounds like a job invented in the last twelve months. In truth it is the oldest job in professional work. Someone comes into the building, watches, builds something, and leaves it running. The carpenter who measures your kitchen. The electrician who wires your house. The accountant who works through your books.
What OpenAI, Anthropic, Palantir, Accenture, and EY are rediscovering in 2026 is that AI needs the same job done. Generic technology, embedded into specific work. Only then does anything happen.
The bottom line for mid-market readers:
- The role exists because foundation model companies admitted AI does not arrive in real work from the outside.
- The economics changed in 2026: build cost collapsed, embedding became the scarce input.
- The gap is real for companies between 50 and 500 employees: needed most, served least.
- The model is replicable as a service for mid-market budgets, not just at half-a-million enterprise scale.
In summary: if you want to know where AI fits inside your own company, the AI Readiness Assessment is an honest place to start.
