Concepts / Turning Custom Solutions Into Reusable Platform Capabilities

Turning Custom Solutions Into Reusable Platform Capabilities

A forward-deployed engineer is assigned to a single customer, embeds with that customer's team, and writes code until the work functions, rather than selling or consulting.

  • Programming

The Gap Between a Product and a Customer

Most software companies build a product at headquarters and then rely on other people to get it working in each customer's environment. A product often does not yet fit what a particular customer needs. When that happens, someone has to close the gap. The question to keep in mind as you read this article is simple: when a product does not yet fit a customer's needs, who closes the gap, and how?

What do you think happens?

When a product does not yet fit a customer's needs, who closes the gap?

  • A salesperson who negotiates a change to the contract

  • A consultant who studies the problem and hands over a report

  • An engineer who lives alongside the customer and whose lessons flow back into the product

  • The customer, who adjusts their own process to match the software

Reveal answer
Answer:

An engineer living alongside the customer closes the gap, and the lessons from that work are then fed back into the product.

The forward-deployed model places an engineer directly inside the customer's team, where the work actually happens, instead of handing the problem to someone far away. The engineer's work then informs the product itself.

What a Forward-Deployed Engineer Is

Definition

A forward-deployed engineer is assigned to a single customer, embeds with that customer's team, and writes code until the work functions, rather than selling or consulting.

The phrase forward-deployed describes where the engineer works. In the early years of the model, being forward deployed was meant quite literally: engineers spent extended periods physically alongside the people who used the software, including in demanding settings. The source is clear, however, that the point of that proximity was not the setting itself. The point was the feedback loop that proximity makes possible.

Key point

The FDE is an engineer whose output is working software inside the customer's environment. The FDE's sole assignment is that one customer, and the assignment continues until the work functions.

Because the assignment is limited to one customer, an FDE is not spread across many accounts. The FDE is not there to win the contract, and the FDE is not there to hand over a report and leave. Those two concerns, winning and reporting, belong to other roles, which the comparison later in this article describes.

Following the Feedback Loop

The model depends on a short feedback loop. An engineer inside the customer's environment sees a problem, fixes it, and tests the fix against real use. The whole cycle happens close to where the problem appears, so nothing has to travel far before it is acted on.

  1. 1
    Observe: the FDE sits inside the room and sees what actually breaks or fails to fit in the customer's real workflow.
  2. 2
    Build: the FDE writes a response, which is code, aimed at that customer's problem.
  3. 3
    Test: the FDE checks the response against real use, not against an imagined version of the work.
  4. 4
    Iterate: the FDE repeats the cycle until the work functions for the customer.

The source contrasts this with vendors that route every issue through a distant queue. Those vendors operate at a much slower cadence. The FDE model exists to remove that delay, and it does so by putting the engineer where the delay would otherwise occur.

The Gravel Road and the Paved Highway

The central image of the model is the gravel road. An FDE builds a gravel road, which is a fast, customer-specific, sometimes rough path from where the product is today to where this customer needs it to be. The road is not elegant, and it is not meant to serve everyone. It exists to get one customer moving.

Product as shippedProduct as shippedGravel road for onecustomerPaved highway in coreproductFDE buildsProduct team generalizes
How does custom, one-off customer code gradually turn into a shared core product path, and what changes from first build to reusable capability?

The diagram shows the two states of the same piece of work. Before, the customer-specific road exists beside the shipped product, built by an FDE for one customer. After, the product team has examined several such roads, found the patterns that recur across them, and converted those patterns into a paved highway. A paved highway is a generalized capability in the core product, one that future FDEs can use without rebuilding it from scratch.

Key point

The gravel roads are the raw experience, and the highways are what that experience becomes once it is generalized. The FDE is responsible for the road, and the product team is responsible for the highway.

How the Two Halves Depend on Each Other

The metaphor captures the heart of the FDE model, and it works because neither half is sufficient alone. Without gravel roads there is nothing to learn from, because the product team has no real customer experience to generalize. Without highways, the same work would be repeated endlessly, with each FDE rebuilding what a previous FDE already built.

Forward-deployedengineerCustomer's teamGravel roadProduct teamPaved highwayCore productembedded withbuildsgeneralizes intobecomes part ofpatterns feed
How are the FDE, the customer's team, and the core product platform connected, and what sits between them?

In this picture, the FDE sits between the customer's team and the work, building roads that carry one customer forward. The product team sits between those roads and the core product, looking across them for recurring patterns. The road is where customer-specific work lives, and the highway is where that work becomes shared.

Building the Capability to Work This Way

The FDE role demands skills that ordinary software work does not always build, so the company eventually created a structured way to produce them. A rotation program called Project Frontline took ordinary software engineers and retrained them as FDEs. The source gives the count of more than 250 engineers who went through it. Its alumni now staff forward-deployed functions at other companies in the industry, including OpenAI and Anthropic.

The existence of a training program is itself evidence of the model's logic. The model depends on engineers who can work inside a customer's team, so the company built a deliberate path to create them rather than relying only on people who happened to have that disposition.

Distinguishing the FDE From Sales and Consulting

The two kinds of work are easy to confuse because both involve code that solves real problems. The FDE role is not a sales job and not a traditional consulting job. Comparing the three roles by what they aim to produce and when their work ends makes the difference clear.

RoleWhat it aims to produceRelationship to the customerWhen the work ends
Forward-deployed engineerWorking software inside the customer's environmentAssigned to one customer and embedded with that customer's teamWhen the work functions for that customer
SalespersonA won contractPursues the dealWhen the contract is won
ConsultantA report handed over to the clientDelivers analysis and then leavesWhen the report is handed over and the consultant leaves

The FDE role differs from sales and consulting in what it produces and when its work is considered finished.

The key difference is the output. A salesperson's work is measured by winning the contract, and a consultant's work is delivered as a report. The FDE's output is software that runs inside the customer's environment, and the FDE keeps working until that software functions.

A Worked Scenario: One Customer's Import Problem

Worked example

From a custom fix to a reusable capability

A generated scenario: a customer's operations team exports order data in a format the core product does not read. An FDE sits with that team and must make the workflow function. Trace how the work moves from a gravel road to a paved highway.

  1. 1
    Observe inside the team

    The FDE watches the operations team perform the export and sees that the core product rejects one column layout that this customer uses.

  2. 2
    Build a customer-specific road

    The FDE writes a custom conversion script that reshapes this one customer's export into the layout the product accepts. The script is fast and rough, and it is meant only to get this customer moving.

  3. 3
    Test against real use

    The FDE runs the conversion on the team's actual exports, finds an edge case with blank rows, fixes it, and runs it again until the workflow functions.

  4. 4
    Hand the pattern to the product team

    The FDE's lessons go back to headquarters. Other FDEs have also built roads that handle inconsistent column layouts, so the product team sees a recurring pattern.

  5. 5
    Generalize into a highway

    The product team converts the recurring pattern into a built-in capability of the core product. Future FDEs can use it without rebuilding the conversion from scratch.

Result

The one-off script for one customer has become a shared capability. The gravel road served its purpose, and the paved highway now serves every future customer with the same need.

From One Customer to the Core Product

The worked scenario follows a pipeline. Each stage hands its result to the next, and the stages run in the same order as the model's logic: customer-specific work comes first, then generalization, then the core product.

FDE writes customcodeCustomer uses it inreal workRecurring patternsfound across roadsProduct teamgeneralizes patternPaved highway in coreproductdeployedobserved across FDEsconvertedshipped
What happens next after an FDE writes custom code for one customer: which steps generalize it, and where does it end up in the core product?

The pipeline ends in the core product, but it does not end the FDE's work. Future FDEs draw on the highway for the next customer, and when they build a road that reveals something new, the cycle repeats. The source describes the model as a continuing loop between the road and the highway, not a one-time handoff.

Mistakes When Reading the Model

  • Common mistake

    Treating the gravel road as the finished product

    What it looks like

    Assuming that because one customer's custom script works, the same script is the answer for the whole product line.

    Why it fails

    The source describes the gravel road as a stage, not a final answer, and says the highway is the product team's responsibility.

    Do this instead

    Treat the road as raw experience and ask whether the product team has generalized the pattern into a highway.

  • Common mistake

    Calling an FDE a salesperson because the role talks to customers

    What it looks like

    Thinking that an engineer who spends time with a customer's team is primarily there to close the deal.

    Why it fails

    The source says the role is not a sales job, and the FDE is not there to win the contract.

    Do this instead

    Judge the role by its output, which is working software inside the customer's environment.

  • Common mistake

    Assuming forward-deployed means only a physical location

    What it looks like

    Believing the model is defined by an engineer sitting at a customer's site.

    Why it fails

    The source notes that physical proximity mattered in the early years, but the point was the feedback loop, not the setting.

    Do this instead

    Focus on the short cycle of observing, building, and testing against real use.

  • Common mistake

    Assuming the FDE's job ends when a report is delivered, which describes consulting rather than the FDE role

Classifying Work as Road or Highway

MEDIUM

A generated scenario: an FDE writes a one-time script that maps one customer's legacy account codes onto the product's new codes. Then the product team notices that three other customers' FDEs wrote nearly identical mapping logic and adds a configurable mapping feature to the core product. Explain which part of this story is the gravel road and which is the paved highway, and say which team owns each part.

Hints
  • Ask whether the work is customer-specific and fast, or generalized and reusable by future FDEs.
  • Look for the moment when recurring patterns across several roads are noticed and converted.
  • The source assigns the road to the FDE and the highway to the product team.
EASY

Explain in two or three sentences why the FDE model needs both gravel roads and paved highways, and what would go wrong if either one were missing.

Hints
  • Consider what the product team would have to learn from if no roads existed.
  • Consider what future FDEs would repeat if no highways existed.

Key Takeaways

  • A forward-deployed engineer is assigned to a single customer, embeds with that customer's team, and writes code until the work functions rather than selling or consulting.
  • The model depends on a short feedback loop in which the engineer observes a problem, builds a response, and tests it against real use.
  • The gravel road is fast, customer-specific work built by an FDE, and the paved highway is the generalized capability the product team creates from recurring patterns across roads.
  • The FDE is responsible for the road and the product team is responsible for the highway, and neither half works well alone.
  • Unlike a salesperson, who aims to win a contract, or a consultant, who hands over a report, the FDE's output is working software inside the customer's environment.