Between May and July 2026, four of the biggest names in AI made the same bet. OpenAI launched a deployment company with more than $4 billion behind it. Anthropic, Blackstone and Hellman & Friedman created Ode, an AI services firm. AWS put $1 billion into a new engineering organisation, and Microsoft committed $2.5 billion and more than 6,000 experts to a unit called Frontier Company. None of these is a new model. All four are ways of sending engineers into customers' offices to make AI work there. The job at the centre of it has a name: the forward deployed engineer. Here's what forward deployed engineering is, why it suddenly matters, and how to get it if you're not a Fortune 500 company.
What is forward deployed engineering?
Forward deployed engineering is a way of delivering software in which engineers work inside the customer's organisation, alongside the people who will use the system, and build it there until it runs in production. The forward deployed engineer (FDE) is not a consultant who recommends, a sales engineer who demonstrates or a support engineer who answers tickets. They are the builder, stationed where the problem is.
The term borrows from military language, where "forward deployed" means stationed close to the action rather than at headquarters. Palantir was using it as a job title by 2009, sending engineers to sit with government and enterprise customers and adapt its software to their data and workflows. For years it was mostly associated with Palantir. By 2026 it had become one of the main ways the AI industry sells.
| Role | What they deliver | What they usually leave to others |
|---|---|---|
| Strategy consultant | Diagnosis, roadmap, business case | Writing and running the software |
| Sales or solutions engineer | Demos and proofs of concept | Production integration and support |
| Agency developer | Software built to a brief | Deciding what the brief should be |
| Forward deployed engineer | Working software inside your processes | Very little: they own the outcome |
Why AI companies are sending engineers into offices
The short version: the models got good enough to stop being the bottleneck, and the bottleneck turned out to be everything around them.
- Most pilots didn't pay off. MIT's NANDA initiative reported in August 2025 that only about 5% of the generative AI pilots it studied achieved rapid revenue acceleration, with most delivering little or no measurable impact on profit and loss (Fortune). The figure has drawn fair criticism, but the pattern behind it is familiar: tools that sit beside the work rather than inside it rarely stick.
- Agent projects are next in line. Gartner predicted in June 2025 that more than 40% of agentic AI projects will be cancelled by the end of 2027, citing rising costs, unclear business value and inadequate risk controls (Gartner).
- Experience is the scarce resource. In the Netherlands, one in six businesses used AI in 2025. Among companies that had considered AI but not adopted it, lack of experience was the most common reason, cited by 73%, ahead of privacy (49%) and legal consequences (42%) (CBS).
Put those together and the business model writes itself. If what holds AI back is not access to models but the skill to fit them into messy, regulated, real-world operations, then the product is that skill, delivered in person.
What the big four launched in 2026
| Company | What it launched | Scale |
|---|---|---|
| OpenAI | The OpenAI Deployment Company, 11 May | More than $4 billion, 19 partner firms led by TPG, and about 150 engineers and deployment specialists from its agreed acquisition of Tomoro |
| Anthropic | Ode with Anthropic, launched in May and introduced under the Ode name on 15 July | A reported $1.5 billion with Blackstone, Hellman & Friedman and others, built on the acquired firm Fractional AI |
| AWS | A Forward Deployed Engineering organisation, 30 June | $1 billion, with plans to embed thousands of experts with customers |
| Microsoft | Microsoft Frontier Company, 2 July | $2.5 billion and more than 6,000 industry and engineering experts |
Sources: OpenAI, TechCrunch, AWS and Microsoft. Google Cloud is hiring for the same role without a new brand: in May it had 59 AI deployment roles posted (Channel Dive). In late July, TechCrunch called FDEs "the AI industry's latest talent obsession" (TechCrunch).
What a forward deployed engagement looks like
Strip away the branding and the vendors describe nearly the same sequence. OpenAI's starts with a short diagnostic of where AI would add the most value, narrows to a few priority workflows chosen with the customer's leadership and operating teams, then builds and deploys production systems inside the organisation. AWS adds an explicit exit: the customer's own engineers progress from watching, to building alongside, to running the systems on their own. In practice it comes down to five steps:
- Go and look. Sit with the people who do the work, not just the people who manage it. The process in the handbook and the process at the desk are rarely the same.
- Choose narrowly. Pick the one to three workflows where volume, time and data line up, and say no to the rest for now.
- Build inside the real systems. Real data, real permissions and real edge cases from the first week, not a prototype on a spreadsheet export.
- Run it alongside the old way. Measure against a baseline taken before the build, so the result is a number rather than an impression.
- Hand over properly. Documentation, training, monitoring and a plan for when the model is wrong.
None of this is new to good engineering. What is new is that the largest AI companies now agree it can't be done from a distance.
What a good forward deployed engineer brings
The title is used loosely, and some critics call it a rebrand of solutions engineering. The label matters less than the combination behind it:
- Production engineering. Integrations, security, monitoring and deployment, not just prompts and prototypes.
- Applied AI judgement. Knowing when a language model, a rules engine or a plain database query is the right tool, and how to test the result.
- Business fluency. Enough grasp of finance, operations or care processes to talk to the people doing the work in their own terms.
- Ownership. Staying accountable until the system is used, not until the demo is approved.
The last one is the real difference, and it's hard to fake. Anyone can be called a forward deployed engineer; nobody can hide whether the system is still running six months later.
The catch: who these programmes are built for
Microsoft's 6,000 experts and AWS's thousands sound like a lot until you spread them across the world's largest companies. The launch clients show where the capacity goes: Microsoft names the London Stock Exchange Group, Unilever and Novo Nordisk. OpenAI's partners, mostly investment firms, between them back more than 2,000 businesses. Ode is the exception on size, aimed squarely at mid-size companies (Ode), but it has around 100 engineers and its private equity backers will steer their own portfolio companies to it (TechCrunch). If you're neither a global brand nor in one of those portfolios, you're unlikely to be first in line.
Two more things are worth knowing before you call one of them:
- Most build on their own stack. OpenAI's engineers connect OpenAI models, AWS deploys into your AWS environment, and Ode is built around Anthropic's Claude. Microsoft describes its approach as multi-model. If you want a model chosen for your data rather than for a vendor's roadmap, that's a different conversation.
- European data questions stay with you. Where the data sits, which processor agreement applies and how the system fits the GDPR (and, in healthcare, NEN 7510) still land on your desk.
None of that makes them bad options. It means the way of working matters more than the logo. You can get forward deployed engineering without being on anyone's strategic account list.
How we run forward deployed engineering at Neurova
We're an engineering company in Eindhoven, and we've always worked to one rule: whoever scopes your system is whoever writes it. Forward deployed engineering is that rule taken to its logical end. An engagement runs like this:
- You tell us the problem. Through the quote form or a short call. Sometimes this is where we tell you AI is not the answer.
- We come and look, when it helps. If the work spans several people or teams, the engineer who will build it spends one to three days in your office, depending on how many teams are involved. We sit with the people doing the work, count volumes and minutes, and look at the systems and data behind each step. What we look for is set out in which business processes to automate with AI first.
- You get a written scope, not a deck. A fixed price and a date for the first system, plus a shortlist of what else is worth doing and what isn't.
- We build inside your environment. Weekly demos on real data, so you see it working, or not working, every week rather than at the end.
- We deploy and hand over. With staff training, monitoring and a documented fallback for when the model is wrong. Then monthly support, or the keys and the documentation. You own the code.
Because we build to healthcare standards, the defaults are strict: European hosting, least-privilege access, audit trails and data minimisation designed in from the first sprint, whether you run a clinic or a logistics company.
Is forward deployed engineering right for you?
It's a good fit when most of these are true:
- The work crosses people and systems. Information arrives by email, gets retyped into one system and checked against another.
- You've tried AI tools and they didn't stick. Staff use ChatGPT on the side, but nothing about how the process runs has changed. Our piece on shadow AI explains why that's a useful starting point.
- The data is sensitive. Customer, patient or financial data shouldn't go into a consumer tool; it needs a system designed for it.
- You don't have an AI team, and don't want to build one yet. Hiring is slow and the skills are scarce. We compare the routes in forward deployed engineers vs AI consultants vs in-house.
It's a poor fit when the problem is a one-off analysis, when an off-the-shelf tool already does the job (we'll tell you if so), or when nobody can spare a few hours to show us how the work is done. And if you've run pilots that went nowhere, read why most AI pilots never make it to production first.
Forward deployed engineering is less a trend than a correction: after years of demos, the industry has admitted that AI has to be built where the work happens. Neurova AI builds custom AI applications, agents and integrations that way, from Eindhoven, for clients across Europe. If you'd like the engineer who would build yours to come and look first, ask for an on-site visit.
Frequently asked questions
What is a forward deployed engineer? A forward deployed engineer is a software engineer who works inside a customer's organisation, next to the people doing the work, to build and ship a production system. Unlike a consultant, they write the code. Unlike a support engineer, they own the outcome until the system runs in daily use.
Why are OpenAI, Anthropic, AWS and Microsoft investing in forward deployed engineering? Because the models stopped being the bottleneck. Between May and July 2026 all four launched units that embed engineers with customers. Research in 2025 had suggested that most generative AI pilots showed little or no measurable impact on profit and loss, and the hard part is fitting AI into real workflows, data and security, which has to happen where the business is.
How is forward deployed engineering different from AI consulting? Consultants typically diagnose and recommend, then hand the build to someone else. Forward deployed engineers diagnose and build: the person who studies your process writes the software, deploys it into your systems and stays until it runs. The deliverable is working software, not a report.
Can companies without an AI team use forward deployed engineering? Yes, and they often have the most to gain. The large vendor programmes start with major enterprises and their backers' portfolio companies, but the model does not need thousands of engineers. A specialist partner can work the same way: time on site to see the real work, a fixed scope, a production build and a proper handover.