This post is for CTOs, operations leaders and procurement owners who have decided to look at forward deployed engineering (FDE: senior engineers who write production code inside your own team and environment) and now have to compare providers who all use the same words.
The main risk is buying a team of contractors under a better name, or a pilot with no path to production. Below we cover the shapes engagements take, how they're priced, what to ask, what to avoid and what to put in the contract. If you need the definition first, start with what forward deployed engineering is.
What are forward deployed engineering services?
Forward deployed engineering services are a paid engagement where a provider places senior engineers inside your team and environment to build and ship a specific production system. The provider is accountable for a working result, and the engagement is designed to end with your team owning it.
A vendor onboarding team configures the vendor's own product, and a consulting project usually delivers recommendations. This is neither. The test is simple: at the end, can your team run, change and extend something that was built in your environment, and do you own it?
What engagement shapes exist?
Most FDE engagements are a single embedded engineer, a pod of 3-5 people, or a short post-launch handoff retainer. The right one depends on how many systems the work touches and how much of your own team's time it needs.
- A single embedded engineer joins your team for a well-bounded workflow. That suits a clear problem in one system, such as automating an intake process or hardening one service. The risk is dependence on one person, so ask who covers absence and how knowledge is shared.
- A pod of 3-5 is a small group with complementary skills: backend, data, integration, and someone who talks to users. It suits work spanning several systems or teams. Deloitte describes pods of about 4-5 people, time-boxed to roughly 3 days to a use case, 3 weeks to a prototype and 3 months to production. Those are one firm's published time-boxes and shouldn't be read as an industry norm.
- The post-launch handoff retainer is a light, short engagement after launch. The provider stays on call while your team takes over, covers incidents, and reviews changes. Treat it as a bridge to your own team.
Our own format is the fixed-scope sprint: 4-6 weeks, a small pod, weekly live demos and production deployment typically within 30-45 days of starting. See forward deployed sprints for how that works.
How are forward deployed engineering services priced?
Five pricing models exist in the market: time and materials, fixed-scope sprints, milestone or outcome-linked fees, retainers, and vendor-bundled FDEs included with a product commitment. Each moves risk differently, so the model tells you who is motivated to finish, and who pays when the work runs long.
- Time and materials means you pay for engineer time. It's flexible, but the provider earns more when the work takes longer, so it needs a firm time-box and weekly demos to stay honest.
- A fixed-scope sprint has a defined problem, a defined window and a defined deliverable. Risk sits mostly with the provider, which means scope has to be tight and agreed in writing.
- Milestone or outcome-linked fees make part of the price depend on hitting an agreed metric. That's attractive, but it only works when the metric is measurable and sits within the provider's reach.
- A retainer gives you a predictable monthly cost for ongoing availability. It's good for post-launch support and risky as a default, because it can quietly turn into permanent staffing.
- Some platform vendors bundle FDE-style engineers with a product or consumption commitment. Those engineers are paid to drive adoption of the vendor's product. We discuss that trade-off in vendor FDEs vs independent FDEs.
We don't quote prices in a blog post, and you shouldn't trust a post that does without seeing your scope. Compare models first, then compare quotes for the same defined scope.
What 10 questions should you ask an FDE provider?
Ask who writes the code, which named engineers are committed, where it runs, who owns it, how it ends, and how success is measured. These ten questions separate real forward deployed engineering from renamed staffing. Use them on the first call, then ask for written answers.
- Do your engineers write production code in our environment, or do they advise while someone else builds?
- Which named engineers will be on our team, and what happens if one is replaced?
- How senior are they, and which production systems have they shipped before?
- Will the system run in our cloud account or private VPC, or on your infrastructure?
- Do we own 100% of the code, database and IP, and is that written into the contract?
- What baseline metric will we agree before work starts, and who measures it?
- How often do we see working software? (Weekly live demos are a reasonable bar.)
- What controls come as standard: access controls, row-level security, audit logging, spend caps on model usage?
- What's the handoff plan, and what does the final week look like?
- Which customer can we speak to about a system that's in production today and not a pilot?
Write the answers down and compare them across providers. We'd treat vague answers to questions 1, 2, 5 and 9 as the clearest warning signs.
What are the red flags of FDE washing?
FDE washing is renaming staff augmentation or consulting as forward deployed engineering. The signs are billing by hours against your instructions, no named engineers, no baseline metric, no handoff plan and no code ownership terms. If those are missing, the title is the only thing that changed.
The pattern has a precedent. Gartner has described "agent washing" in agentic AI: of thousands of vendors claiming agentic features, only about 130 had real ones (Gartner). Hot labels attract relabeling, and FDE is a hot label.
Watch for these in particular:
- The proposal talks about "resources" and "headcount" and never names a system or an outcome.
- The team rotates after a few weeks with no handover.
- Nobody on the call has shipped to production in a comparable environment.
- You're offered a roadmap, a prototype or a report, with implementation priced separately.
- Code lives in the provider's repository or cloud account, and you get access "at the end."
- The engagement has no end date and no plan for your team to take over.
Our side-by-side of the models is in FDE vs consulting vs staff augmentation.
Which contract terms matter most?
Four terms carry most of the risk: named engineers, ownership of code and IP, a handoff and exit plan, and a baseline metric agreed up front. If your contract covers those clearly, many other problems become manageable. If it leaves them vague, expect to argue about them later, when you have the least bargaining power.
- Name the people, their roles, and a notice period or approval process for changes. A senior team swapped for a junior one after signing is a well-known risk in services contracts.
- You should own the code, database and IP from day one, in repositories and cloud accounts you control. At Kaprin the client owns 100% of the code, database and IP, deployed in the client's own AWS or GCP account or private VPC.
- For handoff and exit, specify the deliverables: documentation, runbooks, tests, a walkthrough with your engineers, and a defined end date. Include what happens if you stop early. Forrester argues that FDEs are "training wheels," temporary by design, and that providers should build for handoff and let their footprint shrink (Forrester). Hold providers to it. See handoff by design.
- Record the baseline number before work starts. In our Hunt Digital case, daily reporting took about 4 hours at the start and was cut to about 10 minutes. A before number is what makes the after number credible.
Add access and security terms (who can see production data, how secrets are handled) and a simple way to change or stop after a short first phase.
How do you scope a first engagement?
Pick one workflow with a measurable cost, record today's baseline, name an owner on your side, confirm system access and set a time-box of a few weeks. A narrow first scope produces evidence before you commit to anything larger, and it limits the cost of a wrong guess.
Good first candidates are workflows where people retype data between systems, where a prototype is already in use but fragile, or where a reporting or approval process eats a fixed number of hours every day. Broad mandates such as "AI strategy" or "modernize everything" make poor first scopes.
The provider should be able to restate your problem, baseline and definition of done in a page, within a day or so of the first conversation. Ours is a scope within 24 hours of the first call. If a provider needs weeks of discovery before it can say what it would build, you're probably buying consulting.
If you'd like a second opinion on a proposal you already have, or want to scope a sprint, get in touch. We'll say so if our model isn't the right fit.
