Forward Deployed Engineer vs. Consultant: Why the Handoff Kills Your AI Project
TL;DR
A Forward Deployed Engineer is a senior software engineer who embeds inside a customer's environment, writes production code on their infrastructure, and stays through the first stability milestone after launch. A consultant runs a workshop and exits at the handoff to client engineering. For an AI project, that gap is the difference between an agent running against live traffic and a 60-page recommendation deck nobody acts on. Palantir built its business around this distinction starting in the early 2010s. Anthropic now hires Forward Deployed Engineers at $200,000 to $300,000 base to embed directly with strategic customers. OpenAI, Ramp, Databricks, ElevenLabs, and Google DeepMind run parallel tracks. Hire a consultant when the problem is strategy. Hire a Forward Deployed Engineer when the problem is whether the workflow ships.
What a Forward Deployed Engineer actually does
A Forward Deployed Engineer writes code inside the customer's stack rather than from a vendor sandbox. Palantir invented the role around 2010 and called the title "Delta": engineers who configure a single platform against a single customer's hardest problem, rather than shipping one capability for a thousand customers. The FDE handles data integration into the customer's warehouse, eval harnesses for any LLM workflow in play, prompt iteration with the actual business owner present, and on-call response once the workflow goes live. The artifact is a pull request running against production traffic. The artifact is not a slide deck. When a model upgrade changes output formatting or an upstream API contract shifts, the Forward Deployed Engineer owns the patch. No phase-two handoff is coming.
Why this matters now
The Forward Deployed Engineer role has moved into mainstream hiring over the last 18 months for three concrete reasons. First, LLM-backed agents fail in ways that look fine on a status page: silent retrieval drift, eval scores that regress after a model version bump. None of that is visible to a steering committee meeting once a month. Second, the AI labs took a public position. Anthropic posts Forward Deployed Engineer roles at $200,000 to $300,000 base to embed with strategic customers. OpenAI, Ramp, Databricks, ElevenLabs, and Google DeepMind hire under the same banner. Third, the old consulting model stalled in public. Pilots reach the demo stage and then die at data integration, often after seven figures of spend.
Where FDEs win
Forward Deployed Engineers fix four specific failure modes that a consultant structurally cannot. First, the "we shipped the demo, now what" gap: a typical statement of work ends at the demo, while the FDE engagement begins there. Second, eval ownership for agentic workflows that degrade silently after a model upgrade. The engineer who tuned the prompts reads the regression. Third, customer-specific glue code. Every enterprise has one weird CRM field that breaks routing, and writing the migration SQL requires commit rights, not an advisory seat. Fourth, iteration speed. When the Forward Deployed Engineer sits in the customer's daily standup, the loop from "this is wrong" to "this is fixed" runs in hours rather than the next billable phase.
Where FDEs aren't the answer
The Forward Deployed Engineer model breaks down on commodity work. A standard Salesforce rollout across 40 sales reps is implementation against a known runbook, and a systems integrator with a fixed price will beat a custom engineering team on both cost and predictability. Heavily regulated environments like clinical trials under FDA filing constraints or classified government work often require vendor distance for audit reasons and cannot let an outside engineer touch production directly. Pure strategy decisions are consulting work and should stay there. And when the customer's foundations are missing entirely, with no source-of-truth data and no engineering counterpart on their side, the FDE engagement turns into months of data archaeology. Fix the data layer first, then bring the engineer in.
What to do next
Match the engagement model to the artifact you actually need to leave with. If the deliverable is a strategy memo and a board-ready decision, hire a consulting firm. McKinsey and BCG still own that work. If the deliverable is a live agent that has to keep running after the model version changes next quarter, hire a Forward Deployed Engineer, or a vendor whose entire delivery model is one. Lynk's posture is the second. Our engineers commit code in your repository and stay through the first production incident. They attend your eval reviews and stay paged after launch. The asymmetric reason: a deck cannot drift. A live agent will drift, and the person with shell access is the only one who can patch it.
Want to see Lynk against your own workflow? Book a build session and we'll prototype it in front of you.
Frequently asked questions
What does a Forward Deployed Engineer actually do?
A Forward Deployed Engineer writes and ships production code inside the customer's own infrastructure. The Forward Deployed Engineer owns data integration and eval design for any AI workflow involved and stays on-call after launch. The artifact is a running system rather than a recommendation document.
What's the difference between a Forward Deployed Engineer and a consultant?
A consultant advises and exits at the handoff to client engineering. A Forward Deployed Engineer commits code in the customer's repository and owns production behavior. The Forward Deployed Engineer stays embedded until the workflow is stable and is measured on what actually runs.
Which companies hire Forward Deployed Engineers?
Palantir originated the role around 2010 under the internal title "Delta." Anthropic, OpenAI, Ramp, Databricks, ElevenLabs, and Google DeepMind currently post Forward Deployed Engineer roles or close variants. Anthropic's Applied AI Forward Deployed Engineer postings list $200,000 to $300,000 in base compensation.
When should you hire a Forward Deployed Engineer instead of a consultant?
Hire a Forward Deployed Engineer when the deliverable is a working production system rather than a strategy recommendation. Hire one when the workflow involves LLM agents that need ongoing eval tuning. Use a consultant for everything upstream of that.
How long does a Forward Deployed Engineer engagement typically last?
A Forward Deployed Engineer engagement usually runs from initial scoping through the first stability milestone in production. For an agentic AI workflow, that range is commonly six weeks to six months. The duration tracks the system's stabilization curve rather than a fixed billing phase.