THINK BIG
AI Deployment · Embedded Engineering · Small Business
What Is a Forward Deployed Engineer, and Why Is It AI’s Most Wanted Role? — What Small Businesses Can Do Without an FDE Budget
An FDE works inside your systems and data to wire AI into your actual workflows. Job postings surged more than 800% in 2025, and in 2026 major companies invested billions. AI’s bottleneck has moved from building models to putting them to work.
1. What's happening: AI's hottest job title is an engineer who gets sent out
FDE stands for forward deployed engineer. “Forward-deployed” comes from U.S. military jargon. (Wikipedia, “Forward deployed engineer”) The role was developed by a U.S. data analytics software company. In a 2019 blog post, the company explained the difference: software engineers build the core platform, one capability serving many customers; FDEs deploy the platform to customers, bringing many capabilities to bear on one customer’s critical problems, measuring success by impact on that customer’s goals. The post also addressed a common question: an FDE isn’t a consultant. Consultants typically deliver a one-time analysis or recommendation; FDEs deploy actual software and build long-term solutions together with the customer, doing significantly more engineering. (“Dev versus Delta,” 2019)
Generative AI turned the title into a phenomenon. June 2025: A U.S. venture capital firm called the FDE the hottest job in startups, reasoning that enterprises want AI but need someone to set it up; onboarding an AI agent is no different from onboarding a new employee. (Andreessen Horowitz, 2025-06-04) January–September 2025: According to an analysis by Indeed and the Financial Times, FDE job postings grew more than 800%. (As reported by: Salesforce blog, 2025-11-19) 2026: CNBC reported a major cloud provider investing $1 billion in a new forward deployed engineering unit (CNBC, 2026-06-30), and a major software company committing $2.5 billion and 6,000 employees to a new AI implementation unit (CNBC, 2026-07-02).
In short: AI’s bottleneck has moved from building models to putting them to work. The FDE is the role that grew up in that gap.
2. The debate: real need, or consulting with a new title?
Supporters say FDEs solve the hardest part of AI. Models can be bought and swapped, but every company’s data, workflows, and no-fail rules are different. The same a16z piece points out that even model providers are hiring FDEs: as foundation models become more interchangeable, owning the application layer inside the customer’s workflow matters more. (Andreessen Horowitz)
Skeptics raise three points: “Isn’t this just consulting?” KDnuggets used that as a headline; its answer lands in between: the title may splinter or fade, but the skill of deploying AI into real environments is here to stay. (KDnuggets) It isn’t glamorous. The Wall Street Journal reported in March 2026 that some engineers see it as undesirable because of travel pressure and tight timelines. (Wikipedia) It’s expensive. The a16z piece is literally titled “Trading Margin for Moat.” For large enterprises it’s an investment; for small businesses it’s often out of reach.
Both sides are really describing the same thing: deploying AI takes someone working inside your processes to get the job done. In Taiwan, that’s especially pressing: the country has more than 1.715 million SMEs, over 98% of all businesses, providing nearly 80% of jobs. (Taiwan MoEA, 2025 SME White Paper) Most of them can’t afford a full-time FDE, but they need AI wired into their daily work all the same.
3. What we've learned doing this work: the remote FDE
When Think BIG deploys AI agents for businesses, we do the FDE part: work inside your systems and data to connect AI to your workflows, then test, hand over, and maintain it. The difference is we work as remote FDEs: operating on your machines or cloud environment through secure remote access instead of stationing someone on site. Here’s what we’ve learned, with no client details:
1. Take inventory before building anything. Many companies already have computers, cloud accounts, and some AI subscriptions. We always inventory existing equipment and accounts first, and reuse what works instead of buying twice.
2. Keep assets in your name. Cloud resources, accounts, and data are always set up under the business’s own accounts. After we step back, the system is still yours, and nobody has you locked in.
3. Freeze the scope before work starts. Before we begin, we write down which questions the AI must answer, which data it reads, and what it must never do, and we prepare a set of acceptance questions with known answers. For enterprise projects, delivery happens within 30 days once scope is frozen and data is in place. In our experience, an unclear scope is the most common reason projects drag on.
4. Disable outbound actions during testing. While testing an AI agent, we block its ability to send email, publish posts, or write to external systems, and only turn those back on after its behavior checks out. No “oops, it actually sent that.”
5. Accept on evidence, not on “it’s done.” For example: a local model confirmed to be running on the GPU, answers include citations, and the agent says “I don’t know” when it can’t find something. Each backed by test results, not a demo.
6. Make it visible after handover. Once live, the system is monitored continuously and problems trigger alerts rather than failing silently. The worst failure mode for an AI agent is looking like it’s running when it stopped days ago.
Our biggest takeaway: an FDE’s value isn’t about being physically on site. It’s about someone owning the outcome.
4. Our recommendations: how a small business without an FDE budget can close the gap
1. Pick one workflow before you pick a tool. For example, auto-replies to common customer questions, or search across internal documents. The more specific the goal, the easier it is to verify. 2. Name one internal point person. They don’t need to be technical, but they need to know the process, make decisions, and supply data. 3. Get the scope and acceptance questions in writing. Agree up front on what’s in, what’s out, and which questions will be used to sign off. 4. Keep accounts, data, and systems in your own name. Whoever you hire, make sure what’s delivered belongs to you. 5. Ask how it will be maintained after handover. Who monitors it, how you’ll be notified, and how fast they respond. An AI agent isn’t install-and-forget software.
| Hire a full-time FDE | Remote, embedded-style deployment | |
|---|---|---|
| Best for | Companies running several AI workflows needing long-term in-house development | Small and mid-sized businesses starting with one or a few workflows |
| On site | Yes | No; work happens over secure remote access |
| Scope | Grows with the company’s needs | Frozen up front; next steps evaluated after delivery |
| Watch out for | Hard to recruit; needs someone to manage them | Your point person must supply data and make decisions |
Bottom line: AI deployment isn’t short on models. It’s short on people who wire them into the work. If a full-time FDE isn’t in the budget, find an embedded-style team that owns the outcome.
Remote FDE: inside the workflow, frozen scope, delivered in 30 days, your assets.
(The following reflects Think BIG’s own position and is separate from the neutral reporting above.)
We position ourselves as a remote FDE (embedded engineering) team: delivering the work large companies use FDEs for, in a form small and mid-sized businesses can afford.
Inside the workflow: We don’t hand over a recommendations deck and leave. We connect AI agents to your data and processes until they actually work. Frozen scope, delivery within 30 days: Scope and acceptance questions are written down before work starts, and delivery happens within 30 days once scope is frozen and data is in place. Your assets: Accounts, cloud resources, and data stay in your name, and the model can be chosen to fit your needs and changed later.
Evidence-based acceptance, monitored operation: We test against known answers before handover, then monitor continuously with proactive alerts.
We don’t promise sales results. We promise to wire AI into your actual work and let you see that it’s running.
FAQ
- Q1: What is a forward deployed engineer (FDE)?
- An FDE works on behalf of a software or AI company inside a customer’s systems and data, deploying, customizing, and connecting the product to real business processes. Success is measured by whether the customer’s goal is met, not by delivering a report.
- Q2: How is an FDE different from a consultant?
- According to the company that developed the title, consultants typically provide a one-time analysis or recommendation, while FDEs deploy actual software and build long-term solutions with the customer, doing far more engineering. Some commentators say the line is blurring, but building a working system is the core of the FDE role.
- Q3: Why have FDE job postings surged recently?
- Companies want to use AI but lack people who can connect it to their own data and workflows. According to an analysis by Indeed and the Financial Times, FDE job postings grew more than 800% from January to September 2025, and in 2026 several large cloud and software companies launched dedicated deployment units.
- Q4: How can a small business deploy AI without budget for an FDE?
- Start with one specific workflow, name an internal point person, and work with an embedded-style team that can operate remotely. Get the scope and acceptance questions in writing before work begins, keep accounts and data in your own name, and clarify how monitoring and maintenance will work after handover.
- Q5: Is remote AI agent deployment safe?
- It depends on how it’s done. Confirm that remote access is secure, that accounts and data are in your company’s name, that the AI’s ability to send email or publish is blocked during testing, and that monitoring and alerts are in place after handover. All of this should be agreed before work starts.
This article combines industry reporting with Think BIG’s first-hand practice. All cited official documents and reports are linked; hiring figures reflect each source’s published analysis, and we don’t cite unverified salary figures. Companies are described rather than named in the text, with links in the sources; nothing here evaluates any company or product. Our first-hand experience is shared without client details.