The role started at Palantir in the mid-2010s, scaled at AI companies like Anthropic, OpenAI, and Sierra over the last two years, and has now reached Shopify Plus and ecommerce — where operators with real catalogue depth and operational complexity are hiring FDEs to build production AI systems instead of buying generic agency retainers.
This article explains what an FDE actually does, what they earn, where they work, and why ecommerce is suddenly one of the strongest places to hire one.
The short version
If you’re skimming, here’s the role in one paragraph:
That’s the whole model. The reason it’s growing so fast is that complex software — and especially AI — doesn’t deploy itself. Generic tools fail in specific environments. An FDE solves the specificity problem by living inside it for a quarter or two.
What an FDE actually does, day to day
The job has roughly four phases per engagement, and most FDEs cycle through them every three to six months as they move from one customer to the next.
Embed and scope.
First couple of weeks. The FDE gets read access to the customer’s systems — admin panels, databases, internal tools, analytics. They sit in the customer’s communication channels (Slack, Teams). They talk to whoever owns operations, customer service, merchandising, finance. The goal is to find the two or three workflows where, if software took them over, the compounding return would be largest. By the end of week two there’s an agreed scope in writing. No 90-page deliverable. Just a list of what’s being built, why, and what success looks like.
Build and ship.
Most weeks something ships. The first thing usually goes live by week four — a small piece of software doing real work on real data inside the customer’s environment. From there the FDE layers in additional workflows, integrations into existing tooling, and any custom infrastructure needed. The customer sees demos every week. There are no monthly status reports because the work is visible in production.
Hand off.
The last few weeks of an engagement are about transfer. Documentation. Runbooks. Architecture diagrams. Training the customer’s team to operate the systems. An FDE engagement that creates a black box dependency has failed.
Move on.
The FDE rotates to the next customer. Some firms have FDEs work multiple customers simultaneously; most don’t, because the depth of context required for one customer eats most of the engineer’s attention.
The day-to-day mix is about 70 to 90% coding, with the rest split between customer meetings, reading documentation about the customer’s domain, and writing internal updates. Travel varies — some FDEs are 30% on the road, some are fully remote.
Where the role came from, and where it works now
Palantir invented the modern version of the role in the mid-2010s, calling it the Forward Deployed Software Engineer. The model worked because Palantir’s products needed to be configured deeply for each customer — the software was less a product and more a platform that an FDE would assemble into a custom solution per client.
In 2024 and 2025 the role exploded inside AI companies for the same reason: generic LLMs don’t deploy themselves into enterprises. Anthropic, OpenAI, Sierra, Decagon, Glean, Hebbia, Harvey — basically every applied-AI company you’ve heard of — now hires FDEs in volume. The Anthropic listing for a London FDE in early 2026 paid £225k to £255k base. OpenAI’s London FDE listing currently sits in a similar range.
What’s quieter — and more interesting if you run an ecommerce business — is the same model leaking out of enterprise SaaS into other industries. Forward-deployed engineering for finance, for healthcare, and now for ecommerce. The trigger is the same: companies with real operational complexity realising that generic AI tooling can’t bridge the context gap, and that hiring someone embedded for a quarter is faster and cheaper than building an internal AI team from scratch.
What FDEs get paid
The compensation tells you how the market values the role.
Big AI labs (Anthropic, OpenAI, similar): £225k to £450k+ base, with significant equity.
Big enterprise SaaS (Palantir, Databricks, Snowflake, Salesforce): $200k to $300k base in the US, often with 25 to 50% travel.
Mid-stage applied AI companies (Sierra, Decagon, Harvey, Glean, Hebbia): $180k to $280k base typically, heavier equity than salary, more travel, faster scope.
Independent / consultancy FDEs: Varies enormously. Day rates of £1,500 to £3,000 are common for senior independent FDEs; embedded engagements price as fixed-fee projects, typically £30k to £120k for an 8 to 12 week build.
Why ecommerce operators are starting to hire FDEs
Three things are happening simultaneously, and together they make this the right time for ecommerce-side FDEs.
One — Shopify and AI just officially merged.
Shopify launched its AI Toolkit (an MCP server) on April 9, 2026 — a Model Context Protocol server that connects Claude (and other LLM agents) directly to Shopify’s Admin API. For the first time, an AI agent can do real work inside a Shopify store at the API layer. That’s a structural change. The generic stuff is solved. The specific stuff is wide open.
Two — “AI agency retainers” rarely ship production software.
Most “AI for ecommerce” packages from agencies bolt one or two GPT prompts onto an existing service (“AI-powered SEO”, “AI content automation”) and bill monthly for a workflow that doesn’t compound. The retainer model is the wrong shape for AI work — what you actually want is a piece of software that exists by the end of the engagement, not a recurring service that disappears the moment you stop paying.
Three — Catalogue complexity has gotten unmanageable.
A Shopify Plus store with 2,000 products generates 20,000+ indexable URLs, hundreds of metafields, and tens of thousands of variant-level decisions per quarter. No one is doing this well by hand any more. AI is what makes it tractable. An FDE is what makes it actually ship.
There aren’t many people who can both speak Shopify Plus operations and ship production Claude-based systems. We’re one of them. There’ll be more.
FDE vs agency vs in-house engineer
If you’re an operator considering this, the practical question is which of the three fits.
Agency retainer.
Good for ongoing, predictable work that doesn’t change much month to month — paid media, link building, design refreshes. Bad for engineering work and almost always wrong for AI projects. The retainer model assumes the work is similar each month. Engineering projects ship and finish; they’re poorly served by a model that bills for ongoing presence.
In-house engineer.
The right answer if you have enough sustained engineering work to keep someone busy for a year or more, you can compete on salary against tech companies (a senior AI-fluent engineer in the UK is £100k to £150k+), and you have the technical management capacity to give them a clear roadmap. Most operators don’t.
Forward-deployed engineer.
Good for projects with a defined scope (build the systems that automate X, Y, Z workflows) and a defined timeline (8 to 12 weeks). Better than an agency because you get production code at the end, not a monthly service. Better than an in-house hire when you don’t have the workload to justify a permanent salary, you don’t want to manage technical recruiting, or you want the work shipped fast.
Should your business hire one?
Three questions decide it.
One — Do you have at least one workflow that, if AI took it over, would meaningfully change your operating model? Not “would be slightly faster”. Meaningfully. Examples we’ve seen on real Shopify Plus stores: variant-level merchandising decisions across thousands of SKUs, automated content generation for hundreds of collection pages, customer support classification and routing. If you can’t name one, an FDE is premature.
Two — Do you have someone internally who can be the technical counterpart? Not a senior engineer — just someone who can answer questions about the business, give access to systems, and make decisions about scope. Usually the founder, COO, or head of ecommerce.
Three — Are you OK with a fixed scope and a real budget? A real FDE engagement is £30k to £120k for an 8 to 12 week build. If your AI budget is £2k a month for an agency retainer, the answer isn’t an FDE — the answer is to either grow into a real budget or stay where you are.
If yes to all three, hire one. The compounding return on the systems we’ve built for our own store says it’s the highest-leverage spend you can make on a Shopify Plus business right now.
operator@hollowpoint:~$cat next-step.txt
If you're a Shopify Plus operator considering this.
We’re one of the few consultancies running this model for Shopify Plus operators in the UK. We run our own multi-£m store using the same stack we build for clients — custom Claude agents, MCP servers, Claude Code skills.