Embedded AI engineer (FDE): when your company needs one

Search for "forward deployed engineer" and almost everything you find is written for people who want the job: salaries, interview questions, who is hiring. This guide is for the other side of the table: the company that wants AI running in its operations, is wondering whether it needs an engineer inside its team, and wants to know what exactly it would be buying.

What is a forward deployed engineer, in plain terms?

The term comes from Palantir, which sent its engineers to work at the customer's site instead of building from its own offices. The large AI companies picked it up after ChatGPT. Without the jargon, the idea is simple: the engineer goes where the problem is.

In a mid-sized company that means an engineer who sits in your morning stand-up, asks for access to the ERP, works out why the finance export is missing columns, builds the agent, puts it live, and then sits with the operations team to find out why half of them still do it by hand. That last part defines the role. An embedded AI engineer is judged on whether people use the system, not on whether it was delivered.

The role is growing fast. Postings for forward deployed engineers rose more than 1,000% between January and August 2026 compared with the same period a year earlier, according to Lightcast data reported by Fortune (3 September 2026). Most of that demand comes from the large AI companies hiring FDEs for their own customers. A mid-sized European company is rarely at the front of that queue. The same work can come from an independent engineering firm, which is how we offer it.

What does an embedded AI engineer actually do?

Four things, in this order:

1. Finds the specification with you. The first weeks go on the process as it really runs, the data as it really is and the exceptions nobody wrote down. The specification comes out of that work; it does not come before it.
2. Builds inside your environment. On your stack, under your access rights and security rules, not on a copy of the data on a laptop.
3. Takes the system into production. Logging, monitoring, cost per request, a rollback plan, and an evaluation set (real examples with the correct answer) that shows with numbers whether the system works.
4. Stays until the system is used. Trains the people who will use it, watches where they route around it and fixes the reasons. Then hands it over to one of your engineers, who has worked alongside from week one.

That is the difference between an AI demo and an AI system that runs in your operations. We have written separately about why AI pilots stall before production and what gets them into daily use.

How is it different from a consultant or staff augmentation?

A consultant answers for the quality of the advice. Staff augmentation answers for the hours you directed. An agency answers for meeting a specification. An embedded engineer answers for one outcome: a system in production that your team uses. The skills can be the same. The accountability is different.

The title guarantees nothing. If someone offers you "forward deployed" work and cannot say who answers for the system three months after launch, you are buying consulting or hours under a newer name. The full comparison, with when each model fits, is in AI consultancy vs staff augmentation vs embedded engineer.

When does a company need an embedded AI engineer?

You need the function, not necessarily the title. The model makes sense when all three of these hold:

• you have one important, well-defined use case, but not the team to take it into production alone;
• the system has to live inside your processes and systems, not next to them;
• you want your own people to run and extend it at the end, without the provider.

Many mid-sized European companies are exactly here. Among EU enterprises that considered AI in 2025 and did not go ahead, 70.3% named a lack of relevant expertise, ahead of legal uncertainty (53.6%) and cost (38.4%) (Eurostat, KS-01-26-009, Table 7, published 26 March 2026). And hiring is slow: 57% of EU firms with open IT vacancies reported difficulty filling them in 2024, particularly senior and specialised roles including AI (Eurofound, July 2026).

When is an embedded engineer the wrong choice?

• You do not have a use case yet. Start with choosing your first AI use case. An engineer without a target becomes a very expensive researcher.
• You already have an AI team that only lacks hands. Staff augmentation is simpler, and the result is yours to manage anyway.
• AI is the long-term core of your product. Hire and build the team. An embedded engineer can help you start; it does not replace that team. We compare the options in should I hire an AI engineer.
• The process is about to change. If the workflow will be reorganised next quarter, automating it now means paying for it twice.

What should go into the contract?

The contract is where a real embedded engagement separates itself from consulting under another name. Ask for these in writing before the start:

• A named production milestone: which process, which users and which date count as "running in production".
• An adoption metric with a baseline: how the process performs today, measured in the first week, and the number that shows the team uses the system.
• Something at stake for the provider. If no part of the fee depends on the milestone, you are buying hours.
• Ownership: the code, the documentation and the evaluation set are yours, with no licence needed for what was built for you.
• Pairing: one of your engineers works alongside from week one, and the handover shows they can extend the system alone.
• A named sponsor on your side, with the authority to unblock access and decisions.
• Where the data is processed, and under whose access controls the engineer works.

For high-risk systems, the EU AI Act requires the organisation using the system (the deployer) to assign human oversight to people with the necessary competence, monitor its operation, keep its logs for at least six months and inform workers before it is introduced (Regulation (EU) 2024/1689, Art. 26). For stand-alone Annex III systems these duties apply from 2 December 2027, after the deferral in Regulation (EU) 2026/1744. An engineer inside your team can build them in as the system is built, rather than add them at the end. This is general information, not legal advice.

How does Sapio work as an embedded AI engineer?

At Sapio, a senior engineer works inside your company, on your systems and with your people, and owns one use case until it runs in production and your team uses it every day. Vlad Tudor, our founder, leads every engagement: he sets the target with your sponsor and answers for adoption.

1. First, an audit on your site. We map the processes with the people who run them, check the data on real extracts, pick the use case and tell you in writing whether it is worth doing, including when the answer is "not yet". For most companies this is the Tech Audit, our starting package. For larger groups and regulated firms, it is a longer audit with more than one visit on site.
2. Then, the engineer in your team, three days a week by default. The sponsor, the production milestone and the adoption metric are written down at kickoff, and part of our fee depends on reaching the milestone.
3. Finally, handover. One of your engineers has worked alongside ours from week one. You keep the code, the documentation and the evaluation set.

We do not put an engineer in your team before the audit. If the audit shows a use case is not worth it, we say so, and nothing follows from it.

What we draw on: our team built and runs ai-aflat.ro, the AI assistant for Romanian legislation, which answers from more than 220,000 legislative acts, updated daily. What we build is described on our custom AI agents and AI process automation pages; for companies outside Romania, see nearshore AI engineers in Romania.

Not sure whether you need an engineer inside your team or something smaller? Start with the Tech Call, a free 30-minute conversation: you describe one process and get an honest answer on whether AI belongs in it and what the first step is. Request it here.

Sources

• Fortune, on forward deployed engineers, citing Lightcast data, 3 September 2026
• Eurostat, statistical report KS-01-26-009 on AI use in enterprises, 2026 edition, published 26 March 2026 (Table 7)
• Eurofound, IT sector in focus: evolution of the digital workforce, July 2026
• Regulation (EU) 2024/1689, the AI Act, and Regulation (EU) 2026/1744