A forward deployed engineer is a software engineer who embeds with a customer, learns how the work actually happens, and builds the fix inside the customer's own systems, staying until people use it. A small business can have one too: not as a full-time hire, but by working with a small firm that works this way and bills for the time the problem takes.
If you've read about how the big AI companies sell to large enterprises, you've probably run into the phrase. OpenAI, Anthropic and Google are all hiring forward deployed engineers, and consulting firms are rebranding around them. The idea behind the title is older and simpler than the hype suggests, and it's one of the most useful ways a business of any size can get technology work done.
Where did the forward deployed engineer role come from?
Palantir created the role in the mid-2000s. Its early customers had problems that didn't fit a standard product, and traditional consultants weren't solving them. So Palantir sent software engineers to work directly with the customer, inside the customer's environment, building whatever it took to make the system work in the real world.
Twenty years later, AI companies hit the same wall. A powerful model doesn't run a business process by itself. Someone has to understand the process, connect the systems, handle the exceptions and get people to actually use the thing. OpenAI now runs a forward deployed engineering team with roles in cities from San Francisco to Tokyo, and Anthropic hires for the same work under the title Forward Deployed Engineer, Applied AI.
What does a forward deployed engineer actually do?
A forward deployed engineer is a software engineer who embeds with a customer and owns a problem from end to end. In practice that means four things. They learn how the work really happens, which is rarely how the process document says it happens. They scope the fix with the people who will live with it, not just the person who signed the contract. They build it as production software, inside the systems the business already runs, rather than handing over a design for someone else to implement. And they stay until it works, including the training and adjustments that turn a working tool into a used one. The difference from the usual options is ownership. Because the same person who understands the problem is the one writing the code, nothing gets lost in translation between the people who diagnose a problem and the people who build the fix, which is exactly where most technology projects lose their way.
How is that different from a consultant, an agency or off-the-shelf software?
A consultant hands you recommendations and leaves you to carry them out. A development agency builds to a spec, so it's only as good as the spec you were able to write. Off-the-shelf software does most of what you need and leaves your team doing the rest by hand.
| Option | Who understands the problem | Who builds the fix | Who stays through adoption |
|---|---|---|---|
| Consultant | The consultant | You, after they leave | Usually nobody |
| Development agency | Whoever wrote the spec | The agency, to the spec | Rarely the agency |
| Off-the-shelf software | The vendor, for a typical customer | The vendor, for everyone | Your team, with workarounds |
| Forward deployed engineer | The engineer, from the inside | The same engineer | The same engineer |
Why does the forward deployed model work?
In most businesses, the hard part of a technology project isn't the code. It's knowing what to build.
The problems that cost the most time tend to live between systems: the spreadsheet that feeds the report, the approval that lives in someone's inbox, the data typed into two places, the history stranded in a system you just migrated away from. Nobody owns those gaps. Each vendor's product stops at its own edge, and no one inside the business has the time or technical depth to work across them. We describe that pattern in your best employee is the integration between your systems.
An engineer who sits close to the work can see those gaps, and can fix them without asking you to replace everything you already own.
Why can't small businesses usually get one?
The forward deployed engineers at AI companies serve large enterprise customers on large contracts. A small or mid-sized business is not who they're built for.
Hiring your own doesn't work well either. A full-time senior engineer is expensive, hard to recruit, and often underused. Most businesses under 100 employees have important technology problems, but not a full-time engineer's worth of them every week of the year.
So smaller businesses get stuck choosing between advice they can't act on and software that almost fits. AI-assisted engineering has changed the math on building instead of buying, which we cover in build vs. buy software in 2026.
What does the small-business version look like?
That's the gap Mosaic is built for. We work the way a forward deployed engineer does, sized for a small or mid-sized business:
- We start with the bottleneck, not the technology. We learn how your operation actually runs before recommending anything, and sometimes the right answer is a process change, not software.
- We build inside the stack you already have. Your ERP, your CRM, your document systems, your email. We connect them instead of replacing them.
- We stay through adoption. We train the people who will use what we build, because a tool nobody uses hasn't fixed anything.
- We bill for people's time, not the machines'. Most of our work is hourly, pay-as-you-go. We don't pass through system or AI time in our bills: you'll only be charged for time that real people spent on the task.
In practice, that has looked like an archive for a manufacturer changing ERPs, so years of history weren't lost in the move (more on that in what happens to your old data in an ERP migration). It has looked like one connected place for a law firm's documents, email and practice management system. And it has looked like a cleaned-up CRM for a nonprofit changing how it sells, so leadership, finance and sales could finally work from the same numbers.
How do you know a forward deployed approach is what you need?
It's a good fit when:
- The problem spans more than one system, and no single vendor will own it.
- You've tried off-the-shelf software and your team still works around it.
- You've received advice, but nobody has the time or skills to carry it out.
- You're moving between systems and can't afford to lose what the old one did well.
- You aren't sure whether you need a consultant, a developer or a new product. You just know the process is harder than it should be.
If that sounds familiar, see how we approach custom software and automation on our Software & Automation page. If the question is still which decision to make, rather than what to build, our Advise, Educate & Lead work starts there.


