The Forward Deployed Engineer Era: Why B2B's Hottest New Hire Lives Inside the Customer — and What It Means for the Death of the Traditional CSM

Eight hundred percent.

That's how much hiring for one specific role has surged in the last twelve months — outpacing every other engineering category, every other customer success category, and every other sales category combined. The role barely existed on most B2B SaaS org charts in 2023. Today it is the single fastest-growing job title at OpenAI, Anthropic, Adobe, Salesforce, Databricks, Ramp, Cohere, and roughly every enterprise AI company that has crossed the $50M ARR mark. It pays a median total compensation north of $163,000, with senior bands at Palantir clearing $486,000. New York job postings for it now outnumber San Francisco postings three-to-one.

And almost nobody outside of engineering knows what it is.

It's called the Forward Deployed Engineer — FDE for short — and over the next eighteen months, this single role is going to do something that should keep every Chief Customer Officer awake at night: it is quietly cannibalizing the traditional Customer Success Manager, the Solutions Engineer, the Implementation Consultant, and most of the professional services line item on your enterprise SaaS contract. All of it. At the same time.

For Chief Customer Officers, VPs of Customer Success, Heads of Professional Services, RevOps Leaders, and B2B SaaS Executives, the implications are larger than a hiring fad. The FDE represents the first serious organizational answer to the most expensive open problem in enterprise software: the gap between what a customer signs and what a customer actually gets to use in production. Get this transition right, and your net revenue retention compounds. Get it wrong, and you will spend 2027 explaining to your board why a competitor with a smaller logo count is shipping outcomes you keep promising.

The Pilot Purgatory That Created the FDE

To understand why a role that essentially didn't exist two years ago is now the most-recruited title in enterprise tech, you have to look at a single, brutal number.

95%.

That's the share of generative AI pilots that failed to scale to production deployment, according to MIT Sloan's 2025 enterprise AI research. RAND Corporation's 2025 analysis put the broader number at 80.3% of AI projects failing to deliver intended business value, with 33.8% abandoned before ever reaching production. A March 2026 enterprise survey of 650 technology leaders went further: 78% of enterprises had AI agent pilots, but only 14% had reached production scale. Deloitte's research found that 42% of companies abandoned at least one AI initiative in 2025, with the average sunk cost per abandoned project hitting $7.2 million.

Read those numbers in sequence and one thing becomes obvious: the limiting factor in enterprise AI is not the model. It is not the contract. It is not even the use case. It is the last mile — the hands-on engineering work of wiring a probabilistic system into a deterministic enterprise environment full of legacy data, broken APIs, three-letter compliance acronyms, and at least one VP who insists on running everything through SAP.

Traditional customer success teams cannot do that work. They were never trained to. The CSM toolkit — QBRs, success plans, executive sponsor mapping, NPS surveys — was built for a SaaS world where the product worked the same way for every customer and value was unlocked by improving adoption metrics. The implementation was someone else's problem.

In the AI era, the implementation is the entire problem. And there is no playbook in the CSM canon that solves it.

What an FDE Actually Is (And Why It Breaks Your Org Chart)

The clearest definition comes from Palantir, which invented the model two decades ago and has spent that time refining it into something close to a religion: an FDE is a full-stack software engineer who works directly inside the customer's environment, often for long stretches, building production systems that support real operational decisions. They write code. They architect integrations. They debug production incidents at 2 a.m. They sit in the customer's stand-ups. They have a stake in the customer's outcome that is structurally indistinguishable from being on the customer's payroll.

That last part is where the FDE breaks every existing B2B SaaS org chart.

In a traditional SaaS company, the post-sale handoff goes Sales → Implementation Services → Customer Success Manager → Support → Renewal. Five teams, five sets of incentives, five coordination layers, and a customer who learns to dread the word "handoff." The FDE collapses all five into one technical owner who is empowered to ship code into both your product and the customer's environment.

That is a much more powerful unit of work. It is also a much more expensive unit of work. And the companies pioneering this motion — OpenAI for its largest enterprise contracts, Anthropic for its top deployments, Salesforce for complex Agentforce builds, Adobe for Firefly customizations, Databricks for ML platform integrations — are paying the premium because they have done the math that everyone else hasn't.

Here is the math.

A SaaS company with a 95% pilot-to-production failure rate has a hidden churn problem disguised as a retention problem. The contract renewed; the value never did. Net revenue retention tells you the customer wrote another check. It does not tell you whether the customer's CFO will write the next one. The FDE's job is to convert signed contracts into shipped outcomes inside an 18-month window — before the renewal conversation becomes a referendum on whether the customer ever actually used what they bought.

A 2024 industry analysis found that purchasing AI from specialized vendors succeeds 67% of the time, versus 22% for internal builds. The single biggest variable inside that 67% number is the presence — or absence — of dedicated technical resources owning customer outcomes end to end. That is the FDE wedge.

The Five Things an FDE Does That a CSM Structurally Cannot

If you only remember one thing from this article, remember this list. Every line item below is something a customer expects to be solved post-sale in 2026 and that no traditional CSM organization is staffed to solve.

1. Build the integration that wasn't in the SOW. Every enterprise deployment surfaces an integration the sales team didn't know existed. The CSM's only move is to file a feature request. The FDE writes the integration in a customer-specific repo, ships it to production within the quarter, and graduates the most-requested patterns into the core product roadmap. Customer-driven product development, with the customer literally in the room.

2. Fix the customer's data before fixing your software. Enterprise data is the dirtiest secret in B2B SaaS. CSMs hand a customer a CSV template and hope. FDEs run schema migrations, write transformation logic, and stand up the data pipelines that make your product viable in the customer's environment in the first place. Roughly 60% of FDE time at AI companies is spent on data work that has nothing to do with the AI.

3. Pass the security review that's blocking the rollout. A typical enterprise AI deployment now sits behind 217-question security questionnaires, model risk management reviews, data residency proofs, and at least one auditor who wants to see your evals. CSMs route these to a beleaguered InfoSec team. FDEs answer them in-line, often by writing the controls the auditor is asking about.

4. Translate model outputs into operational workflows. This is the work that separates "we have a chatbot" from "we replaced 40% of Tier 1 ticket volume." Intercom's Fin scaled from five customers to 7,000 in eighteen months, achieving a 67% resolution rate — and the documented driver was the FDE motion that built customer-specific routing, fallback, and escalation logic on top of the core product. CSMs cannot do this. Solutions engineers cannot do this in scale. Only an embedded engineer can.

5. Loop production reality back into the roadmap. This is the FDE's secret strategic value. Every line of code an FDE ships in a customer environment is a real-time signal about what your product is missing in the wild. Companies that run a tight FDE-to-product feedback loop ship the right roadmap eighteen months earlier than companies that rely on win/loss interviews and CSM call notes.

Look at that list and try to identify which item is a CSM's job. None of them are. They were never going to be.

The Quiet Restructuring Already Happening Inside Your Peers

While most B2B SaaS leadership teams are still debating whether to add an FDE pod, a quiet restructuring is already underway at the companies that figured this out twelve months ago. The pattern is consistent enough to be a template.

Stage one: a single FDE is hired into engineering, reporting to a VP of Engineering, and assigned to two strategic accounts. Renewal expansion on those accounts beats the company average by 30 to 70 percent in the next twelve months.

Stage two: the FDE is cloned into a small pod — typically three to five engineers — and the company starts assigning them to the top 10% of accounts by ARR. The internal language begins to shift. "Implementation" becomes "deployment." "Onboarding" becomes "go-live." "Customer success plan" becomes "outcome contract."

Stage three: the company splits its post-sale function in two. Long-tail accounts continue to be served by a leaner CSM organization focused on adoption metrics, training, and renewal motion. Strategic accounts — typically the top 20% that drive 80% of revenue — move under an FDE-led structure where every account has a named technical owner. Professional services as a P&L line begins to disappear; the FDE org absorbs its mandate.

Stage four: the FDE function becomes the source of net new product. The roadmap process explicitly designates a percentage of engineering capacity (commonly 20 to 30 percent) for "FDE-validated patterns" — features that have already shipped in three or more customer environments and are now ready to be productized.

By the time a company hits stage four, its competitors are still arguing about whether to upskill their CSMs.

The Customer Success Identity Crisis

This transition is going to be uncomfortable for an entire profession, and it would be dishonest not to say so.

Customer Success as a discipline emerged in the early 2010s in response to a real problem: subscription software companies were churning customers because nobody owned the post-sale outcome. Gainsight, Totango, ChurnZero, and a generation of CS leaders built a category around the insight that adoption needed a home. That insight was correct, and it remains correct for the long-tail of every B2B SaaS book.

But the work has bifurcated. For the strategic top of the customer pyramid — the accounts where AI complexity, integration depth, and operational embedding actually live — the work is now technical. Coaching a customer's executive sponsor through a renewal is no longer the unlock. Shipping the missing eval framework is.

The strongest CS leaders are already responding. They are hiring engineers. They are repositioning their function as a hybrid GTM-engineering org. They are negotiating headcount transfers from professional services. The weakest CS leaders are doing what every threatened function does: writing memos about why their existing motion still matters.

If you are running a customer success org in 2026, the only question that matters is which of those two leaders you want to be when the next QBR cycle starts.

The Practical Playbook: What B2B Leaders Should Do This Quarter

Reorganization is the wrong place to start. Hiring is the wrong place to start. Vendor selection is the wrong place to start. Here is what to do this quarter, in order, with an honest pace.

Map the deployment gap on your top 20 accounts. For each, list (a) what was sold, (b) what is in production today, (c) what specifically is blocking the gap. Three-quarters of the blockers will be technical. That is your business case.

Hire your first FDE inside engineering, not customer success. Reporting line matters more than title. The early FDEs need to see themselves as software engineers who happen to spend time in customers, not as CSMs who learned to code. Anchor them to a VP of Engineering or a CTO. Pay them at the engineering market rate — which, per current data, means a base in the $150K to $200K range and total comp that lands between $200K and $400K depending on stage and geography. Underpaying for this role is the single most common failure mode.

Pick two pilot accounts that hate you and two that love you. The instinct is to deploy FDEs only at happy customers to build proof. Resist it. Half the value of the FDE motion is converting at-risk accounts. The other half is establishing the playbook on accounts where you have permission to take risks.

Define the FDE's success metric as customer outcome, not utilization. The fastest way to break this motion is to hold FDEs accountable for billable hours or implementation milestones. Hold them accountable for time-to-first-value, expansion ARR per account, and named outcomes shipped per quarter. The Palantir version of this is the most disciplined: an FDE is graded on whether the customer renewed at higher commit and whether a feature shipped in their environment was upstreamed into the core product.

Build the productization loop early. Without a structured way for FDE work to flow back into the core product, you will end up running a high-priced consulting firm inside your software company. The most successful FDE orgs run a monthly cadence where engineering leadership reviews the top FDE-built patterns and explicitly votes on which ones graduate to general availability. This is also the meeting where your roadmap stops being aspirational and starts being earned.

Communicate the change to the existing CS team early and honestly. The worst version of this transition is the one where a CSM finds out from a job posting that their account is being moved under an engineering-led structure. The best version is one where the CSM is given a real choice: a path into FDE-adjacent technical roles, a path into the long-tail CS function with refreshed scope, or a path out with dignity. Pick the best version.

The Strategic Window Is Closing Faster Than the Org Chart

Here is the part that should sharpen your timeline. Forward Deployed Engineering is currently a competitive advantage because most B2B SaaS companies haven't built it yet. That window will close — fast.

OpenAI is already assigning FDEs to its highest-value enterprise customers. Anthropic has built an in-house forward-deployed engineering function. The Big Four consulting firms are importing the Palantir model wholesale to monetize the next phase of enterprise AI. EY has launched UK and Ireland FDE roles specifically positioned for AI deployment. The category is professionalizing.

In twelve to eighteen months, "We deploy FDEs on every strategic account" will go from being a differentiator on a sales call to being a baseline expectation on every enterprise RFP. The companies that have already built the muscle will be selling outcomes. The companies that haven't will be explaining themselves.

There is a version of 2027 in which the B2B SaaS companies that figured this out look back at this moment as the inflection point — the quarter where they shifted from selling licenses to selling outcomes, from staffing CSMs to staffing engineers, from running pilots to shipping production. There is also a version in which they look back and realize that their net revenue retention quietly leaked away to the competitor who treated the gap between contract and outcome as the only thing that mattered.

The FDE is not a hiring trend. It is the answer to the question every enterprise customer has been asking, in every renewal conversation, for the last three years: "You sold us this. Why isn't it working?"

Whoever builds the function that can answer that question wins the next decade of B2B SaaS. Everyone else is going to spend 2027 wondering where their pipeline went.