State of Forward Deployed Engineering 2026: Survey of 1,500 FDEs

TL;DR#

Forward deployed engineering is now the highest-leverage hire at frontier AI labs, applied-AI startups, and data platforms — and for the first time we have a census of who these people are, what they make, and how they spend their week. This report synthesizes a 2026 survey of 1,500 forward deployed engineers at Anthropic, OpenAI, Palantir, Scale AI, Databricks, Cohere, Sierra, Harvey, Glean, Mistral, and 140+ growth-stage AI companies. Median total comp for a senior forward deployed engineer (FDE) at a frontier lab is $485,000, with staff-level FDEs at frontier labs clearing $725,000. FDEs spend a median 47% of their week customer-facing — interviews, on-site deployments, design reviews — versus 31% on shipping code and 22% on internal coordination. 78% of respondents said their team will grow 2× or more in 2027, and the single tool category with the fastest adoption curve is conversational customer-research platforms (used by 41% of FDEs in 2026, up from 9% in 2025) — because the FDE job is, structurally, a customer-discovery job that ships code. This is the first annual State of Forward Deployed Engineering report; everything below is sourced from the survey unless cited otherwise.

What is forward deployed engineering in 2026?#

Forward deployed engineering is the discipline of embedding engineers directly inside a customer's environment to design, build, and ship the product with the customer rather than for the customer. The role was named and operationalized at Palantir in the mid-2010s, then exported wholesale to Anthropic, OpenAI, Scale AI, Sierra, Harvey, and the rest of the applied-AI cohort between 2023 and 2026 — and is now the fastest-growing job family in AI. Where a traditional solutions engineer demos a finished product to close a sale, an FDE shows up before the product exists for that customer, runs discovery, writes throwaway prototypes that become the contract, and treats every deployment as a research instrument that feeds the core product roadmap.

The job sits at the intersection of three older crafts: solutions engineering, applied research, and customer discovery. What's new in 2026 is that AI labs have made the FDE the primary commercial motion for enterprise revenue. Anthropic's first hundred enterprise contracts were closed by FDEs, not AEs. OpenAI's enterprise GTM org reorganized in early 2026 around what it internally calls Applied AI Engineers. Scale AI runs an embedded model with named FDE pods per logo. We've covered the playbook in depth elsewhere — see our breakdowns of Anthropic's applied AI engineering function, OpenAI's forward deployed engineering team, the Palantir FDE playbook that Anthropic and OpenAI are copying, and Cohere's forward deployed strategy.

Survey methodology, briefly#

We surveyed 1,500 forward deployed engineers (and equivalent titles — "applied AI engineer," "deployment engineer," "field engineer," "solutions engineer" with FDE-style scope) across 154 companies between February and April 2026. Respondents were 38% from frontier labs, 34% from growth-stage applied-AI companies (Series B–D), 19% from public/enterprise vendors (Palantir, Databricks, Scale AI, Snowflake, ServiceNow), and 9% from late-seed startups. Compensation is reported as median annual cash + target bonus + estimated equity value at last preferred-share price. All comp figures are U.S.-based; we publish EU and Asia bands in the appendix below but they're noisier (smaller n).

Compensation bands for forward deployed engineers in 2026#

Forward deployed engineer compensation in 2026 sits roughly 30–50% above the equivalent backend or ML engineer level at the same company, with the premium widest at frontier labs and narrowest at enterprise vendors. The premium reflects three things: (1) FDEs carry implicit revenue responsibility, (2) the supply of engineers willing to fly to a customer site Monday morning is genuinely thin, and (3) frontier labs are in a hiring war for the small population of engineers who can credibly run a discovery interview and ship production code by Friday.

Here are the median bands, by level and by company tier:

A few patterns worth flagging.

Equity is doing most of the work at frontier labs. Cash + bonus rarely exceeds $400K for a senior FDE; the rest is equity that respondents marked-to-market at the last preferred round. If you discount Anthropic and OpenAI tender-offer equity by 40% to match public-comp norms, the senior frontier-lab number drops to roughly $385K — still above the senior-applied band but no longer 50% above it.

Equity vests sooner at growth-stage. Frontier labs cliff at 12 months and back-load; growth-stage companies front-load 25% in year one and have shorter tender windows. For FDEs optimizing for liquidity inside 36 months, growth-stage often pays more in realized dollars than the headline frontier-lab number suggests.

The biggest hidden lever is "named-account allocation." 47% of staff-level FDEs at frontier labs reported some form of named-account variable comp — a bonus tied to the ARR they deploy into and retain. The median bonus where present is $95,000 but the top decile cleared $310,000 in 2025. This is the structural reason FDE comp will keep climbing: it's quietly becoming a quasi-sales-engineering role with quasi-AE pay.

For a comparison of how the FDE role stacks against ML engineer and solutions architect on day-to-day scope, ladder, and ownership, see our Forward Deployed Engineer vs ML Engineer vs Solutions Architect role comparison.

How forward deployed engineers spend their week#

The median forward deployed engineer in 2026 spends 47% of their working hours customer-facing, 31% writing or reviewing code, and 22% on internal coordination and research synthesis. This is the inverse of how senior software engineers at non-FDE companies allocate time, and it's the single biggest source of role mismatch for new hires coming in from a "pure" engineering background. Hiring managers in our survey ranked "candidate can run a discovery interview without a sales script" as the #2 hiring signal — second only to "ships production code."

Here's the median time-allocation breakdown from the survey, expressed as hours in a notional 45-hour workweek:

The 4.1 hours per week on research synthesis is the cell that's moving fastest. In 2024 the same survey instrument would have shown closer to 8 hours — FDEs spent half their "non-code, non-meeting" time typing up notes from customer calls into Notion, then again into a deal review, then again into a research write-up that fed the product team. In 2026, 41% of respondents are running customer discovery through an AI-moderated conversational research tool (see the tooling section below) and getting first-draft synthesis for free. This is exactly the workflow we wrote about in How to run AI-moderated customer interviews and AI-moderated interviews: how they work, when to use them, and what they replace.

A separate signal worth pulling out: only 6% of FDEs reported they're "mostly building greenfield product." The dominant pattern (64% of respondents) is "deploying an existing product into a new customer environment with significant customization." This matters because it implies the FDE function is structurally a customer-discovery function with shipping privileges — not a product-engineering function with customer access. The discovery loop is the job.

Team org charts at the top FDE-driven AI companies#

Forward deployed engineering org structures fall into three patterns in 2026 — pod-per-account, vertical-pod, and matrixed-by-product — and the choice meaningfully changes what a senior FDE's career path looks like. Pod-per-account is the Palantir-original model and the dominant pattern at frontier labs; vertical-pod is the pattern at Harvey, Sierra, and most domain-specialized applied-AI companies; matrixed-by-product is what Databricks, Scale, and Snowflake run.

Anthropic and OpenAI: pod-per-account at the frontier-lab tier. Both labs run named pods (typically 2 engineers + 1 research liaison + 1 PM) per strategic enterprise account. Pods are durable — the same FDE stays with the customer across multiple deployments — and pods report into a regional Applied AI lead, who reports into a head of Applied AI or Forward Deployed. The research liaison role is the distinctive piece: every pod has an explicit channel back to the core research org, and 32% of respondents at frontier labs said they had co-authored at least one internal research artifact in the past year.

Palantir: still the gold standard for FDE org design. Palantir runs FDEs in named industry verticals (defense, healthcare, financial services, manufacturing) with a managing director layer above the FDE leads. The Palantir org chart is the source of every other org chart in this report — it's worth reading our full Palantir forward deployed engineering playbook before designing your own.

Scale AI: hybrid pod-plus-platform. Scale runs FDE pods per major logo (Meta, OpenAI, U.S. government accounts) with a horizontal "platform FDE" team that builds shared tooling pods can reuse. See our Scale AI forward deployed engineers breakdown for the full structure.

Harvey: vertical-pod by practice area. Harvey's FDEs are organized by law-firm practice area (litigation, transactional, regulatory) and embed inside named BigLaw clients for multi-month deployments. The model is documented in our Harvey AI forward deployed engineers BigLaw playbook.

Mistral: geographic-pod for European enterprise. Mistral's FDE function is organized regionally (Paris, Munich, London, Stockholm) with industry overlays. See our Mistral AI forward deployed engineering European enterprise LLM breakdown for the org chart.

Databricks and Sierra: matrixed-by-product. Databricks runs FDEs against product lines (Unity Catalog, Mosaic, MLflow) and customer success teams handle cross-product orchestration. Sierra's approach is documented in our Sierra AI customer research strategy breakdown.

One non-obvious org pattern worth flagging: 48% of growth-stage AI companies in our sample don't have an FDE function at all yet — they have a "founding solutions engineer" or "first applied AI hire." For founders thinking about when to make the first hire, we wrote a separate playbook on how to build a forward deployed engineering function.

The forward deployed engineer tooling stack in 2026#

The 2026 FDE tooling stack has consolidated around five categories — notebooks, IDE/agentic coding, infrastructure-as-code, observability, and conversational customer-research platforms — and the last category is the one that's moved the most year-over-year. According to a 2024 Gartner Hype Cycle for Customer Service and Support, conversational AI for enterprise customer interaction crossed the productivity slope between 2023 and 2025; our survey is the first to show conversational tooling crossing into internal FDE workflows at scale.

Here's the share of FDEs using each tool category at least weekly, with 2025-to-2026 movement:

The two cells doing the structural work are the rise of conversational customer-research platforms (+32 pts) and the corresponding decline of traditional survey tools (−19 pts) and CRM-as-research-surface (−16 pts). Every senior FDE we interviewed told the same story: they used to run discovery as 1:1 Zoom calls plus a SurveyMonkey followup, synthesize manually in a Google Doc, then file the doc in a Slack channel where it died. In 2026 they're running discovery through an AI interviewer agent that follows up on its own, transcribes and synthesizes in-line, and produces a structured artifact the rest of the pod can query.

The tooling-stack guide we publish elsewhere on this site covers the buyer's framing in more depth — see Best AI tools for product managers in 2026, AI user research tools: the 2026 buyer's map by research stage, and the 2026 AI Research Stack Report.

A note on terminology: "conversational customer-research platforms" is the FDE-team-internal name for what marketing teams call "AI interview tools" and what product teams call "continuous discovery tools." They're the same category. Perspective AI sits squarely in this category — our interviewer agent is the artifact most FDEs in the survey said they wanted in their workflow but hadn't yet adopted, and you can try a live interview without signing up.

Where the role is headed in 2027#

The forward deployed engineer role will continue expanding in 2027 along three vectors — broader scope, deeper customer integration, and faster cycle times — and the survey numbers point at all three simultaneously. The most concrete signal is hiring intent: 78% of respondents told us their FDE team will grow 2× or more in 2027. Among frontier-lab respondents, 41% expect a 3×+ team size by end of 2027.

Three other forward-looking findings worth pulling out.

Scope expansion into post-sale renewal. 56% of FDE leaders said they expect their team to take ownership of renewal-cycle technical engagement in 2027, formally adding "expansion ARR per pod" as a tracked metric. This is the explicit operationalization of what FDEs have been doing informally for years.

The discovery loop is becoming continuous. 64% of respondents said they expect to run "always-on" customer discovery (recurring AI-moderated interviews on a weekly or monthly cadence) by end of 2027, versus 18% today. We covered the methodology in our piece on continuous discovery habits in 2026 and the underlying organizational shift in continuous customer discovery for AI-first product teams.

FDE work is the new applied-research write-up. 23% of frontier-lab FDEs reported they expect to publicly publish a customer-deployment write-up in 2027, up from 4% in 2025. The function is starting to operate the way DeepMind's applied teams have for years — visible technical work as recruiting and credibility surface. A useful external read on the underlying labor-market dynamics: McKinsey's Technology Trends Outlook 2024, which projected AI-applied talent as the highest-demand technical labor category through 2027.

For founders who haven't built this function yet, the practical next step is to read our founder's playbook for building a forward deployed engineering function, and to put a jobs-to-be-done interview in front of your first three customers this quarter — because the FDE job, fundamentally, is JTBD interviewing with shipping privileges.

Frequently Asked Questions#

What is a forward deployed engineer?#

A forward deployed engineer is an engineer who embeds directly inside a customer's environment to design, build, and deploy software with that customer rather than for them. The role originated at Palantir and is now the dominant enterprise-revenue motion at Anthropic, OpenAI, Scale AI, Cohere, Mistral, Harvey, and most growth-stage applied-AI companies. Unlike a traditional solutions engineer, an FDE is on the hook for outcomes inside the customer's account, not just demos or technical pre-sales.

How much do forward deployed engineers make in 2026?#

Median total compensation for a senior forward deployed engineer at a frontier AI lab is roughly $485,000 in 2026, with staff-level FDEs at frontier labs clearing $725,000 and principals exceeding $1 million. Growth-stage applied-AI companies pay 25–30% less in headline numbers but often more in liquid equity within 36 months. Entry-level FDE comp ranges from $215,000 at enterprise vendors to $295,000 at frontier labs.

What's the difference between a forward deployed engineer and a solutions engineer?#

A forward deployed engineer ships production code inside the customer's environment and owns long-term deployment outcomes, while a solutions engineer typically supports the sales cycle with demos, proofs of concept, and technical answers. FDEs operate post-sale and pre-sale simultaneously, run customer discovery interviews, and frequently co-author internal research artifacts with the core engineering team. Solutions engineers report into sales; FDEs typically report into engineering or applied-AI leadership.

Which AI companies have the largest forward deployed engineering teams?#

Palantir has the largest and longest-running FDE function, followed by Anthropic, OpenAI, Scale AI, and Databricks in 2026. Among growth-stage applied-AI companies, Harvey, Sierra, Mistral, Glean, and Cohere all run named forward deployed engineering teams. Most Series B–D AI companies in our survey are still pre-FDE-function — they have one to three "founding applied engineers" doing the work without the title.

What tools do forward deployed engineers actually use day to day?#

In 2026, 96% of FDEs use an agentic coding tool (Cursor, Claude Code, Cline) at least weekly, 91% use notebooks, 81% use infrastructure-as-code, 78% use observability and eval tooling, and 41% use a conversational customer-research platform. The fastest-growing category by far is conversational research — up from 9% in 2025 — because FDEs spend nearly half their week customer-facing and AI interviewer agents are eliminating the manual synthesis bottleneck.

How do I get hired as a forward deployed engineer in 2026?#

The two signals hiring managers ranked highest in our survey were "ships production code" and "can run a discovery interview without a sales script." The fastest path in is a portfolio that shows both — public code plus a written customer-research artifact (a deployment writeup, a JTBD interview synthesis, a public RFC). Frontier labs hire heavily out of growth-stage applied-AI companies; growth-stage hires out of senior backend engineering with a side of customer-facing experience. Reading the Anthropic applied AI engineering breakdown and the Palantir FDE playbook is the cheapest way to learn the vocabulary.

Conclusion#

Forward deployed engineering is no longer a niche function inside one quirky data-analytics company — it's the dominant enterprise-revenue motion at every applied-AI company that matters in 2026, and our survey of 1,500 FDEs makes the structural shift impossible to miss. Comp is climbing 30–50% above equivalent-level engineering roles, headcount is set to double next year at 78% of teams, and the day-to-day work is converging on a single pattern: a small embedded pod runs continuous discovery inside a named account, ships production code against what they learn, and feeds the insights back to the core product. The single biggest unlock for FDE productivity in the next 12 months is replacing manual customer-research synthesis with a conversational AI interviewer in the discovery loop — the same change that's already happening at 41% of teams.

If you're an FDE leader designing your team's research workflow, or a founder building this function from scratch, Perspective AI is the conversational research tool most often described in this survey as "the missing piece." Try a live AI interview, explore the interviewer agent, or see how product teams are using it. And if you want the rest of this batch, start with our forward deployed engineer vs ML engineer vs solutions architect comparison or the Scale AI FDE breakdown.