Embedded team, staff aug, fixed SOW, or hire?
Hire, augment, contract a fixed scope, or embed a pod. Every option puts engineers on the work. They differ on what it costs, how soon something ships, and who holds the system in 18 months.
Four models, the same six questions.
Put them on one page before you sign anything.
| Question | In-house hire | Staff augmentation | Fixed SOW | Embedded pod |
|---|---|---|---|---|
| What you buy | Permanent headcount | Hours or seats | A scoped deliverable | An outcome, plus embedded capacity |
| Who scopes the work | You | Your managers | The vendor’s PM and your sponsor | The same engineers who ship it |
| Time to first shipped change | After a 4–6 month hiring cycle, for one client | Days to weeks of résumés | Weeks of SOW negotiation first | About 2 weeks to assemble, first PRs by Day 5 |
| Cost | $195K–$220K a year per senior hire, fully loaded, in one client’s market | Time and materials or monthly seats, set by the vendor | A fixed fee for the scope: agents from $8,000, the Sprint at $25,000 | $3,000–$7,500 / engineer / mo |
| Who holds it at month 18 | Your team | Whoever is still staffed | Often nobody; the engagement closed | A designed handover, or managed operations |
| Classic failure | Seats that stay open | Code you rewrite, and no AI depth | A demo that never reaches production | Sold as a body shop, with no evals or runbook |
Gigabit figures are our published prices. The hiring figures are one client’s, from the HR-tech case study, not a market average. Staff augmentation rates vary by vendor, so none are quoted.
Each is right somewhere.
Hire in-house
Right whenAI is core to your product, the work never stops, and you can actually fill the seats.
Wrong whenThe roadmap is already slipping while the requisition sits open.
Staff augmentation
Right whenYou have a working AI roadmap and senior reviewers, and the backlog is blocked on hands, not judgment.
Wrong whenNobody on your team can evaluate an AI engineer’s work. Then you have bought a second management problem.
A fixed SOW
Right whenOne workflow is clear, success is measurable, and you want a flat fee with a finish line.
Wrong whenThe statement of work ends at a demo and assumes “your team will operate it” without naming who.
An embedded pod
Right whenYou need senior AI engineers now, the work spans more than one system, and you expect to own it eventually.
Wrong whenThe need is one small, well-defined automation. Buy that as a fixed-price agent instead.
Embed against hire, in real numbers.
A Series B HR-tech company had two senior seats open for four months.
After 200+ applications, 40 phone screens and three lost offers, they embedded three Gigabit engineers instead. The pod cost $14,400 a month. Sprint velocity went from 34 points to 78. Read the case study.
The test for any vendor, us included: can they name who scopes the work, who ships it, and who is accountable after launch? If not, you are buying staff augmentation, whatever the brochure calls it. And if you have a working roadmap and only need hands, staff augmentation is the right purchase.
Embedded, augmented, contracted or hired
What is the difference between an embedded AI team and staff augmentation?
Staff augmentation sells capacity: people on tickets that you manage. An embedded team is a named senior pod that scopes, ships and stays accountable for the outcome inside your sprint board, Slack and repositories. If the vendor cannot name who owns evals, the runbook and post-launch operation, it is staff augmentation with better marketing.
Is an embedded AI team the same as forward-deployed engineers?
At Gigabit, yes. Forward-deployed describes how the engineers work: inside the client’s systems and process rather than over a wall. Embedded AI Teams is the name of the engagement you buy.
How much does an embedded AI team cost compared with hiring?
Embedded AI Teams are $3,000–$7,500 per engineer per month, with a 2–4 week trial and no lock-in. For one HR-tech client, three embedded engineers cost $14,400 a month, or $207,600 a year, against about $600,000 for the equivalent US hires, with first pull requests in 5 days instead of after a 4–6 month hiring cycle.
When should you choose a fixed SOW instead of an embedded team?
Choose a fixed SOW when one workflow is clear, success is measurable and you want a flat fee with a finish line: for example a Gigabit agent from $8,000 flat, or the $25,000 two-week Transformation Sprint when you still need the diagnostic. Choose an embedded pod when the work spans several systems and you need capability to transfer, not a single deliverable.
Who should own the AI system 18 months after launch?
Decide that before you sign. Staff augmentation leaves ownership with whoever is still staffed. A fixed SOW often ends when the engagement closes. An embedded pod is designed to hand over to your team or to managed operations at $3,000–$20,000 a month. If the honest answer is “nobody yet”, do not buy a build-only statement of work.
Write down who owns it in 18 months.
Tell us the work and that one line. We’ll say which model fits, including when it isn’t ours.