Forward Deployed Engineering
Forward Deployed Engineering for Mid-Size Businesses: How the Model Scales Down
Forward deployed engineering was built for governments and giant enterprises. Here's what changes when a mid-size business uses it, and how to tell if you're ready.
Forward deployed engineering (FDE) is a way of building software where engineers work inside the customer's business, beside the people who do the work, instead of building from a requirements document at a distance. Palantir started it in the mid-2000s for governments and very large companies. In 2026, OpenAI, Anthropic, Microsoft and Google all hire forward deployed engineers, and nearly all of that effort is pointed at the largest organisations in the world. This article is about the other end of the market: what forward deployed engineering looks like for a mid-size business, and when it is worth paying for.
What is forward deployed engineering?
Forward deployed engineering means the engineers who build your AI system sit with your teams while they build it. They watch how the work really happens, build into the systems you already run, ship to one team, measure whether people use it, and keep improving it until they do. The engineer who finds the problem is the one who writes the code.
The role has gone from niche to mainstream quickly. Fortune reported Lightcast data showing forward deployed engineer job postings up more than 1,000% from January to August 2026 against the same months of 2025. In India, Indeed data put the growth at nearly 19 times between June 2025 and June 2026. The demand is real. The question for a mid-size business is whether any of it is meant for them.
Why the model was built for large enterprises
The original FDE model assumes a big contract. A vendor places a small team of senior engineers inside one customer for months, and the same Lightcast data put the median advertised salary for the role at more than $188,000. Those economics only work when the customer is a government agency, a bank or a global manufacturer, so that is who the vendors staff for.
Even a16z, which helped popularise the term, warned in January 2026 that most companies copying Palantir's playbook risk becoming expensive services businesses. Their point was about startups trying to sell the model. The consequence for buyers is simpler: most mid-size businesses never meet a forward deployed engineer, because they are not the customer the model was priced for.
Why mid-size businesses need it more, not less
The irony is that the problems FDE solves are sharper in a mid-size business than in a large one.
There is no internal AI team to hand over to. A large enterprise can buy a platform and assign a team to make it work. A mid-size business usually has a small IT team that is already busy keeping the ERP, the network and the laptops running. If the vendor leaves at go-live, nobody picks the tool up.
The real process isn't written down anywhere. In a mid-size business, much of the work lives in WhatsApp groups, inboxes, spreadsheets and the heads of a few experienced people. A requirements document written in a meeting room captures the official process. The engineer who spends a week with the people doing the work captures the real one.
A failed pilot costs more than money. MIT's NANDA initiative reported in 2025 that 95% of generative AI pilots it studied produced no measurable return, and pointed to brittle workflows and poor fit with daily work as the cause. In a business of a few hundred people, staff remember the last tool that was switched off. The next one starts with less goodwill.
Decisions are closer. This one works in your favour. In a mid-size business, the owner or COO who approves the work is often a short walk from the people who will use it. The FDE loop of observe, build, ship and adjust runs faster when a decision takes a day rather than a quarter.
What changes when forward deployed engineering is scaled down
The principles stay the same. What changes is the size of each step, so that the first commitment is small and you only continue if the previous step delivered.
It starts with a fixed-scope audit, not a contract. An AI Audit for a growing business takes one to two weeks, remote or on site. It maps where AI and automation would pay off and ends with a ranked list and a recommendation for what to build first.
It builds one workflow at a time. Not a platform, and not a transformation programme. One workflow, with one team, measured against success criteria you agreed before the build started.
The engineers are remote-embedded most of the time. Daily working sessions with your team, and on site when the work needs it, such as discovery on a factory floor or training at a branch.
The people you meet are the people who build. There is no layer of solutions consultants between you and the engineers. The person who sat with your dispatch team is the person who writes the code that changes it.
It doesn't end at go-live. Once the first workflow is in daily use, a dedicated engineer on a monthly retainer keeps improving it from user feedback and takes the same loop to the next workflow.
What a mid-size engagement looks like, step by step
Find. The engineers spend time with the people who do the work, from the front line to the back office. They map every path the data takes, including the ones nobody admits to, such as the spreadsheet that gets emailed every Friday. They measure a baseline per team: time spent, how late the system of record is updated, data quality and who actually uses the current tools.
Build. The first workflow is built into the stack you already run: your ERP, CRM, helpdesk or inbox. Where nothing you run fits the work, it is built from scratch. Error handling, monitoring and audit logs go in from the first version, and a person approves anything the system is unsure about.
Ship. It goes live with a single team, in real work, usually within weeks of the audit. Problems show up in one team before they show up in ten.
Adopt. This is the step most vendors skip. Usage is measured per team, per person, per week. When a record goes stale, a nudge goes out. If the nudge is ignored, the right manager hears about it.
Iterate. Every week the tool changes based on what users say and what the usage data shows. Once it is the default way of working, the loop moves to the next team or workflow.
Signs your business is ready, and signs it isn't
You are probably ready for forward deployed engineering if:
- A workflow repeats every day and eats hours of skilled people's time, such as re-keying orders, chasing updates or assembling reports by hand.
- Your system of record is always out of date because updating it is a chore.
- You can give the engineers real access to the people doing the work, not just to managers.
- Someone senior will look at the usage numbers every week.
It is probably the wrong model if:
- An off-the-shelf tool already solves the problem for a small monthly fee. Buy the tool.
- You want a strategy document or a vendor shortlist rather than something built and running.
- You are about to replace your ERP or reorganise the team the tool is for. Wait until the ground stops moving.
- Nobody can spare time to be shadowed. Forward deployed engineering without access to the work is just remote development with extra travel.
How Agilemorph works with mid-size businesses
Agilemorph is one of India's first forward deployed engineering firms, with more than 40 projects delivered and engineers from Amazon and YC-backed startups. For growing businesses we run the same loop described above, at the smaller scope: an AI Audit first, then a first build that is live within weeks. You can see the engagement model in full on How we work.
One current example: we are embedded with a Dubai-based owner with factories in Africa, where our engineers work directly with factory users to find where AI fits into the systems already running. It is still in progress, so we are not quoting results. That is the point of the model: we report what the usage data shows, when it shows it.
FAQ
What is the difference between a forward deployed engineer and a consultant?
A consultant studies your business and hands over recommendations. A forward deployed engineer studies your business, writes the code, ships it to production and stays until people use it. The finding and the building are done by the same person, so nothing is lost in a handover.
Can a mid-size business afford forward deployed engineering?
The mid-size version is designed so the first commitment is small. It starts with a fixed-scope audit, then one workflow with one team. You decide whether to continue at the end of each step, based on what the previous one delivered, rather than signing a long contract upfront.
Do forward deployed engineers have to work on site?
No. Most mid-size engagements are remote-embedded, with daily working sessions with your team. Engineers come on site when the work needs it, such as discovery on a shop floor or training a team where they work.
How long until the first workflow is live?
Usually within weeks of the audit. The first workflow goes live with one team first, so problems surface early and the rollout to other teams is based on real usage rather than a demo.
Who owns what a forward deployed engineer builds?
You do. The code, workflows, documentation and data belong to your business, and the system runs on your infrastructure, in your cloud or on premise.