Services · Engineering

Forward deployed engineering: we stay until your AI works in production

Forward deployed engineering is a delivery model where senior engineers work inside your environment, with your real data, systems and users, until an AI system hits an agreed production goal. AMT formalised its Forward Deployment Engineering methodology in 2024. You agree the outcome before we start, and we price the work around it, not around hours.

Schedule Technical Discovery

Your first conversation will be with a senior technical partner, not a salesperson.

Why this model exists

Most AI systems work in a demo. Far fewer work inside a real company, where data is messy, permissions are strict, systems are old and users do unexpected things. The gap is not the model. It is integration, access, evaluation and cost at real volume. Forward deployed engineering puts engineers where that gap is, inside your environment, and keeps them there until the system works.

In 2026 the model went mainstream. OpenAI, Anthropic, AWS and Microsoft each launched or funded forward deployed engineering units to embed engineers inside customers. We track what that means for buyers in our forward deployed engineering insights.

What we agree before day one

An FDE engagement fails or succeeds on five things agreed up front. We write all five down before any engineer starts.

What we agree before day one
Agreement	What it means	Example
The production goal	One measurable result that means "done"	A claims workflow running end to end for one team, at an agreed accuracy
Access within days	Systems, data and credentials the engineers need, with a date	Read access to the document store and a test SSO group in week one
One decision owner	A person on your side who can say yes or no fast	The operations lead for the workflow
What the system must never do	Actions, data and users that are off limits	No customer-facing replies without human approval
The handover	What your team must be able to run and change when we leave	Runbooks, evaluation suite and access transferred

How an engagement runs

Discovery, two to three weeks. We map the gap between the demo and production and write the five agreements.
Embed. Engineers join your environment, tools and rituals.
Connect to reality. Data sources, internal APIs, single sign-on and permissions.
Make it trustworthy. Retrieval, guardrails and an evaluation set built from your own questions and data.
Production ready. Monitoring, cost and latency checks, rollback and runbooks.
Prove and hand over. The production goal is measured, then the system passes to your team, or AMT runs it.

Each stage moves through the delivery gates on How we work. The rule that runs through all of it: no blind AI-generated code, and a named human accountable at every gate.

Forward deployed engineering vs the alternatives

Forward deployed engineering vs the alternatives
	Forward deployed engineering	AI consulting	Staff augmentation
What you get	A system working in production	A strategy and recommendations	Engineers working on your backlog
Who owns the outcome	AMT, against the agreed goal	You	You
Writes production code	Yes, every day	Rarely	Yes
Priced on	An outcome agreed up front	Days or a fixed fee	Hours or a monthly fee
Ends when	The production goal is met and handed over	The report is delivered	The contract ends

The full comparison with staff augmentation is in this guide.

What our engineers build during an engagement

RAG pipelines that answer from your documents (RAG development)
AI agents that act across your tools (AI agent development)
MCP servers that give models controlled access to your systems (MCP server development)
Conversational and voice AI for customers and staff
The integration, identity and monitoring around all of it

AMT has built AI systems since 2013, starting with IBM Watson. See what we have shipped on Case studies.

Is forward deployed engineering right for you?

Yes, if:

The AI already works somewhere: a demo, a pilot or another customer.
The blocker is inside your environment: data, integrations, permissions or users.
Your side can give access and decisions within days.

No, if:

Nothing works yet. Start with AI development.
You want slides, not shipped code.
You need engineers on your backlog under your direction. Hire a forward deployed engineer or use a dedicated team.

FAQ

What is forward deployed engineering?

A delivery model where engineers work inside the customer's environment, with real data, systems and users, until a system hits an agreed production goal. It closes the gap between a demo that works and a deployment that works.

How is forward deployed engineering priced?

AMT prices it around an outcome agreed before the work starts, not by the hour. Across the market it is rarely billed as time and materials: a GoodFirms survey of 133 companies in mid-2026 found only 3.6% do so.

How long does an FDE engagement last?

Until the deployment hits its production goal. It starts with a two to three week discovery that sets the goal and the milestones.

What do you need from us?

Access to systems, data and the people who run the workflow, within days, plus one decision owner and a production goal you are willing to sign.

What is the difference between forward deployed engineering and staff augmentation?

With staff augmentation you direct engineers on your backlog and own the outcome. With forward deployed engineering AMT owns delivery against an agreed production goal.

Is this only for AI?

It is built for AI deployments, where the gap between demo and production is widest. The same model works for any system that has to fit a complex real environment.

Tell us what works in the demo, what breaks in your environment, and what production should look like.

Want one engineer instead of an engagement? →