Put a contractor's day rate next to a salaried engineer's monthly cost and the contractor looks outrageous — which is why this comparison, done naively, misleads more budgets than any other in engineering. A salary is not the cost of an employee; it's the most visible line in a stack that includes employer contributions, recruiting fees, months of below-full productivity, management and tooling overhead, and the amortized risk that the person leaves and you pay the whole entry price again. Augmentation has its own stack — the rate itself, onboarding time you pay for, the provider's margin inside the rate. Neither model wins universally. The honest answer is a crossover: augmentation tends to win on shorter horizons and uncertain needs, in-house tends to win on long, stable ones — and the crossover point sits later than the naive rate-vs-salary glance suggests.
The full cost stack of an in-house hire
The fully loaded cost of an employee stacks several layers on top of gross salary, and every layer is real money or real risk. Exact percentages vary by country and company, so treat the following as the checklist of what to count, not as universal constants.
- Gross salary — the visible line, and the only one most comparisons use.
- Employer contributions and benefits — social charges, payroll taxes, insurance, pension, equipment and perks. Commonly adds a meaningful double-digit percentage on top of gross, varying widely by jurisdiction.
- Recruiting cost — agency fees or the internal equivalent: sourcing, dozens of interview-hours from your senior engineers, and the vacancy months while the search runs.
- Ramp-up — new hires take months to reach full productivity in your domain; you pay full price from day one for a fraction of the output.
- Management, tooling and space overhead — licenses, hardware, office or stipend, plus a slice of a manager's finite attention.
- Attrition risk, amortized — if an engineer might leave within a few years, a share of the entire recruiting-plus-ramp entry price belongs in each year's true cost. This is the line naive comparisons always omit.
The augmentation cost stack
Augmentation's stack is shorter and mostly visible, which paradoxically makes it look worse: everything is in the rate, staring at you on one invoice. The rate has to cover the engineer's pay, the provider's employer burden, non-billed time between engagements, the vetting and matching machinery, replacement guarantees, and the provider's margin. On top of the rate, count your side honestly too: onboarding days you pay for before full productivity, a ramp discount for the first weeks, and utilization — you pay for the committed time whether your backlog feeds it well or not. What's absent matters just as much: no recruiting fee, no vacancy months, no severance exposure, and no amortized attrition risk — if the person leaves, replacement is the provider's contractual problem, not a five-figure restart.
- Rate × committed time — the whole visible cost; includes provider margin and the engineer's effective employer burden.
- Onboarding and ramp — days to a couple of weeks of paid time before full-context productivity; real, but far shorter than employee ramp for well-matched seniors.
- Utilization discipline — a paid engineer waiting on access requests or an empty backlog is pure waste; this cost is yours to control.
- Exit and replacement — largely priced in: notice periods are short and replacement is typically the provider's obligation.
A worked example, side by side
The numbers below are deliberately round and purely illustrative — actual salaries, employer-cost percentages and rates vary enormously by country, seniority and market conditions. Run the same structure with your own real numbers; the structure, not the figures, is the takeaway. Scenario: one senior AI engineer needed for a 12-month build, starting from a standing start (no candidate in hand).
| Cost line | In-house hire (illustrative) | Staff augmentation (illustrative) |
|---|---|---|
| Gross salary / rate | €100,000 salary | €700/day × ~220 days = €154,000 |
| Employer contributions & benefits | +€25,000 (illustrative 25%) | Included in rate |
| Recruiting cost | +€20,000 (fee or internal equivalent) | €0 |
| Vacancy before start | ~3 months of nothing shipped while searching | Days to ~2 weeks to start |
| Ramp to full productivity | ~3 months at partial output, salary fully paid | ~2 weeks at partial output |
| Attrition risk, amortized | A share of the ~€45k entry cost per year of expected tenure | Provider's problem — replacement contractually covered |
| Rough 12-month total | ~€145,000 cash — for roughly 6-7 productive months in year one | ~€154,000 cash — for roughly 11+ productive months |
The crossover logic: when each side wins
Read the illustrative table carefully and the real pattern emerges: in year one, augmentation frequently wins outright — similar cash, far more productive months, because the search and ramp costs land entirely on the in-house side. From year two onward, the in-house line drops (no recruiting, no vacancy, full productivity) while the augmentation line stays flat, and hiring pulls ahead on pure arithmetic — provided the person stays and the need persists. That 'provided' is the whole decision. Duration certainty is the main variable: a need you're confident lasts three-plus years favors hiring; a need that might end, pivot or change skill-shape within 18 months favors augmentation, because you never pay the entry price for capacity you stop needing. The hybrid is often optimal: augment now to start shipping this month, hire deliberately in parallel without deadline pressure, and let the augmented engineer help ramp the hire — converting the overlap from double cost into a knowledge handover.
- Under ~12 months of need, or genuine uncertainty about the need: augmentation wins on math and on risk.
- Multi-year, stable, well-defined need with a healthy local hiring market: in-house wins from roughly year two onward.
- Urgent start plus long-term need: augment immediately, hire in parallel, plan the handover explicitly.
- Redo the math at each renewal — a contractor extended by default for three years is usually the sign nobody re-ran the numbers.
The non-financial factors that should override the math
When the arithmetic lands within perhaps twenty percent either way, stop doing arithmetic — the qualitative factors dominate at that margin. Knowledge retention: an employee's accumulated context compounds for years; a contractor's leaves at the end unless you engineer the transfer deliberately (pairing, documentation, ramping your own people alongside). Team fabric: permanent teams build trust and shared standards that pure contractor rotations struggle to replicate. Cutting the other way — hiring-market reality: if the search for a genuinely senior AI engineer in your market takes six-plus months, the spreadsheet's 'hire' column is fiction, and the actual choice is augmentation now versus nothing for two quarters. Optionality has value the spreadsheet omits: in a fast-moving field, paying a premium to be able to change skill-shape in a month is often worth more than the premium. And flexibility cuts both ways in a downturn — reducing contractor capacity is a notice period; reducing employees is severance, morale damage and legal process.
