Staff augmentation vs managed services in one table
Most of the disappointment with either model comes from buying one while expecting the other: hiring augmented engineers and expecting them to own the result, or buying a managed project and expecting to steer it week by week.
| Staff augmentation | Managed services | |
|---|---|---|
| You buy | Capacity: engineers on your backlog | An outcome, against a written scope |
| Who directs the work | You: your priorities, your definition of done | The vendor, against the agreed scope |
| Who carries delivery risk | You | The vendor, up to the edge of the scope |
| Priced by | Time: per engineer per month | Scope: a fixed price per result |
| Our price | $4,000–$7,000 per engineer per month, three-month minimum, then rolling | Quoted fixed-price; most engagements $15,000–$120,000 |
| Time to start | 1–2 weeks from signing | A written scope within 48 hours of the discovery call |
| What you need on your side | A product owner and a tech lead with hours every week | Someone who can sign off a scope and accept the result |
| Changing direction | Free: you re-prioritise the backlog | A change to the scope, so a change to the price |
| Ending it | With notice, after the minimum | At delivery, with the runbook handed over |
| Fails when | Nobody on your side has time to direct it | The scope is still moving while it is priced |
Staff augmentation, outstaffing and team extension are one purchase under three names. A dedicated development team is the same model at a larger scale: two to eight engineers plus a tech lead working only on your product, billed the same way. Every role we staff like this.
What “managed services” means in this comparison
The phrase covers two different purchases, and vendors rarely say which one they mean.
- A managed project. The vendor builds a defined result — a feature, a product, a migration — at a price fixed against a written scope, and hands it over. This is what we sell as managed delivery.
- Managed operations. The vendor runs a system or a function for you on an ongoing basis — monitoring, incidents, patching — under a service agreement. This is the classic managed service provider model.
Our managed delivery ends in one of two ways, and you choose which before the build starts: support after launch — monitoring, incident response, scaling and iteration — or a full handover to your own team, with the runbook delivered alongside the code. Either way, augmentation is the opposite purchase: the operations stay with you because the direction does.
What each model really costs
The headline numbers are not comparable on their own, because each model hides a different cost on the buyer’s side.
Staff augmentation is the monthly rate plus your own time. $4,000–$7,000 per engineer per month is about $25–$44 an hour at 160 hours — the arithmetic is shown in what an AI engineer costs. What does not appear on the invoice is the product owner who writes the tickets, the tech lead who reviews the work and the weeks of onboarding into your codebase. That onboarding is why our minimum is three months: below that, it eats most of the value and both sides know it.
Arithmetic, not a quote. Three engineers for three months is 9 engineer-months, so $36,000–$63,000 at the published monthly rates. A fixed-price LLM feature added to a product you already run is $15,000–$30,000 over 2–4 weeks; a full AI product $40,000–$120,000 over 8–12 weeks. The ranges overlap, which is the point: price does not decide between the models.
Managed services are the fixed price plus the cost of a scope that moves. The number holds as long as the scope does. Every change after signing is a change request, and a project whose direction changes weekly turns a fixed price into a negotiation. What you save is management time: you spend it once, on the scope and on acceptance, instead of every week.
When staff augmentation is the right call
- You have a product owner and a backlog, and the gap is hands. The engineers join your board, your standups and your review process; your definition of done stays yours.
- The scope will keep changing. Iteration on a live product, where next month’s priorities depend on this month’s numbers, is the work a fixed price describes badly.
- The knowledge has to stay in your company. Augmented engineers work in your repository from the first commit, and you hold 100% of the IP from day one.
- You need people this month, not after a hiring cycle. We can start within 1–2 weeks of signing.
When managed services are the right call
- The result can be written down. A feature with acceptance criteria, an MVP with a defined first release, an integration with a known system on each end.
- Nobody on your side has the hours to run a team. Added capacity without direction drifts, and drift is expensive at senior rates.
- You need a number for the budget before the work starts. A written scope with a fixed price against it arrives within 48 hours of the discovery call.
- The work is outside what your team does. A one-off build in a stack you will not keep staffing for is cheaper to buy as a result than to manage as a team.
Moving from one model to the other
The two models are not a one-way choice, and the most common healthy pattern uses both in sequence.
- Managed first, then augmented. The first version is bought as a result because it can be written down. Once it is live and the work turns into iteration, engineers who know the code join your team on a monthly basis. This is cheap when the runbook and the code are handed over with the first version, not negotiated after it.
- Augmented first, then managed for a slice. A team already working on your backlog finds a self-contained piece — a migration, an integration, an AI feature — that can be written down. That slice is scoped and priced separately while the team carries on with everything else.
The thing that makes either switch cheap is the same: ownership of the code from the first commit, and a codebase the next team can read without the previous one in the room.
Five questions that decide it
- Can you write the result down, with how you will accept it?
- Who on your side will prioritise the work every week, and do they have the hours?
- Will the scope still be true in a month?
- Does the knowledge need to stay in your company after the work ends?
- Do you need a fixed number for the budget before anything starts?
Yes to the first and fifth, no to the second: managed. Yes to the second, third or fourth: augmentation. Yes to everything usually means the first version is managed and the iteration after it is augmented.
For AI work specifically
The same split holds, with one addition: an AI result can only be written down once there is an evaluation set that says what a correct answer is. With one, an AI feature or an AI MVP is a good fit for a fixed price. Without one, the scope cannot be accepted, and the honest model is AI engineers on your team while the definition of “correct” is worked out. The AI-specific risks, the running cost of a production system and the questions to ask a vendor are in AI development outsourcing.
We sell both models: managed AI and software delivery at a fixed price, and engineers on your team billed per month. Every band we publish.