A Forward Deployed Engineer (FDE) is a senior engineer who works inside a client’s team to find where time is lost and build AI solutions that fit the team’s daily work.
Every company has work that repeats every week. A legal team checks each new contract against the same list of clauses. A support team answers the same questions by digging through the same shared drives. Someone in finance rebuilds last month’s report by hand. Most of these teams already have AI licenses, but few have changed how they work.
In Deloitte’s State of AI in the Enterprise 2026 survey of business and IT leaders, about 60% of workers had approved AI tools, but only 30% of organizations were redesigning key processes around AI.
Who’s a Forward Deployed Engineer?
A Forward Deployed Engineer (FDE) is a senior engineer who works inside a client’s team to find where time is lost and build AI solutions that fit the team’s daily work. The role exists to close that gap.
That makes the role different from a consultant, who usually ends with a recommendation, and from a solutions engineer, who usually helps a customer set up a vendor’s product. An FDE starts from the team’s problem, builds the fix, and puts it into daily use.
The role has grown fast. Job postings for Forward Deployed Engineers reached 5,330 in April 2026, according to Indeed data Business Insider reported. Most of what is written about the role treats it as a career move. This guide looks at it from the other side: whether your company needs one.
When does a company need a Forward Deployed Engineer?
Repetitive work is the usual starting point, but not the only one. You do not need all four signs below. Two or three are usually enough to start the conversation.
1
A team loses hours to the same kind of work, week after week.
2
Nobody has mapped how that work gets done. Writing the steps down is part of an FDE’s job, so the team does not need a process document first.
3
The team is too close to the work to see the waste. People inside a process rarely see where their own time goes. An outside view often shows it.
4
The team can spare a few hours a week to answer questions, explain what it does, and check the FDE’s findings.
If none of these ring true, an FDE is probably not the right first step. The next section covers what usually fits better.
When is a Forward Deployed Engineer the wrong choice?
An FDE is a focused, hands-on role. In these cases, something else will serve you better.
The tool is already chosen and only needs setting up. If the decision is made and the job is configuration, the vendor’s own onboarding team will usually be faster.
The goal is to change how your engineers write code. That is a different process, with its own path. An FDE works on the work around the code, such as specifications, reviews, and handovers.
The whole company has to change at once. That needs governance, internal champions, and a phased rollout. It calls for an AI transformation program, not one embedded engineer.
Nobody will own the result after handover. An FDE leaves behind workflows the team has to run. If no one can take them over, the work fades once the engineer leaves.
What can a Forward Deployed Engineer build?
An FDE makes sense when the fix for a team’s problem has to be built rather than bought or learned. That usually means connecting AI to company data, writing the prompts and checks a workflow needs, or fitting it into the systems people already use.
| Use case | What it looks like today | Where it shows up | What an FDE might build |
|---|---|---|---|
| Document review | Contracts or invoices checked by hand against the same list | Legal, finance, procurement | A checking workflow that flags only the exceptions for a person |
| Repeated writing | Proposals and reports started from a blank page | Sales | A drafting workflow fed with the team’s own templates |
| Scattered knowledge | Answers buried in tickets and shared drives | Support, operations | An assistant that answers from company documents and cites them |
| Triage | Requests and emails sorted and forwarded by hand | Support, IT | Routing that sorts each request and drafts a first reply |
| Data between systems | Copy and paste between CRM, spreadsheets and ERP | Sales operations, finance | Integrations that move and reconcile the data |
| Recurring reports | The same monthly report rebuilt by hand | Finance, management | Automated data pulls and a first-draft summary |
How does a Forward Deployed Engineer project work?
Infinum has spent 21 years building software for other companies. The most useful thing that experience brings is judgment about which problems deserve a build at all.
An FDE project runs in three stages.
1
Observing the work.
The FDE will sit with the team, follow the requests and documents as they move, and find the hours that disappear into repetitive manual work.
2
Comparing the options.
For each problem, the FDE will weigh a feature you already pay for, an off-the-shelf product, an open-source model on your own servers, and a custom build.
3
Building and handing over.
The FDE will put the best option into daily use, document it, and leave the team able to run and scale it.
What an FDE finds can be bigger than one workflow. Sometimes the analysis shows that the real fix is a new product or a customer-facing AI feature. At that point, the FDE’s findings become the brief for an AI-native product team, so nothing learned in the first weeks is lost.
We tested the approach on our own HR team first. An FDE learned how hiring, scouting, and CV preparation worked day to day, and built AI workflows that now run them.
And for our client A1, an FDE embedded with the engineering team found that writing technical specifications took the most time. The FDE built an AI workflow around that one task, using tools the team already had.
Vendor FDE or independent FDE?
Forward Deployed Engineers typically come from two kinds of companies. Many are employed by companies that sell an AI model or software platform, and are sent to help customers roll out that product. Others work for an independent partner that your company hires directly.
In research published by Cursor in July 2026, rebuilding the same software with four model setups cost from about $1.3k to $10.6k, depending on which model handled which step.
The difference is that a vendor’s engineer knows one product well, which helps if that specific product is chosen. But no single AI model leads on every task. In research published by Cursor in July 2026, four model setups rebuilt the same software and passed the same test suite. Costs ranged from about $1.3k to $10.6k, depending on which model handled which step.
Infinum does not sell any AI platform. We are partners of several model and cloud providers, so we can recommend whichever fits the task. Workflow logic stays separate from any single model, so your team owns the code, prompts, and documentation.
| Vendor FDE | Independent FDE | |
|---|---|---|
| Employed by | A company that sells an AI model or platform | A partner your company hires |
| Starts from | Where the product fits | Where the team loses time |
| Options considered | The vendor’s own product | Whatever fits, including tools you already own |
| Strongest when | The product is already chosen | The right tool is still an open question |
| What you keep | A deeper setup in one product | Workflows your team owns and can move between models |
The answer starts with one team
The quickest way to find out whether your company needs an FDE is to test one team. Pick the team that loses the most hours to repetitive work, and count them before anyone builds anything. Five people losing four hours a week to the same report adds up to more than 1,000 hours a year. You do not need AI licenses or a process document to start.
Check that team against the four signs. If most of them fit and the number is large, give an FDE a handful of its problems and six to eight weeks.
Check that team against the four signs. If most of them fit and the number is large, give an FDE a handful of its problems and six to eight weeks. Then judge the result by how many of those hours came back, and what the team did with them instead.
If one of your teams fits, see how we approach AI transformation, from a single team’s workflow to company-wide change.
