The category of supplier you shortlist matters more than which firm inside that category wins. One senior engineering day in Europe costs somewhere between €150 and €2,500 depending on the category, and almost all of that spread is explained by three things: who absorbs estimation error, how many juniors sit under each senior, and whether you get named people or a logo.
The five categories
| Category | What you are actually buying | Indicative senior day rate, Europe | Breaks when |
|---|---|---|---|
| Big-4 style consulting practices | Advisory plus delivery, board-level cover, audit-grade documentation | €1,400–2,500 | The job is six months of engineering with no governance question attached |
| Global systems integrators | Managed scope at volume, multi-year run contracts, a single throat to choke | €700–1,400 blended | You need four strong engineers rather than forty mixed ones |
| Offshore delivery centres | Cost per seat, shift cover, large and repeatable workloads | €150–350 | The work needs daily iteration with your product owner |
| Nearshore specialist providers | Named senior engineers on time and materials | €400–700 | You want a third party to guarantee a fixed outcome |
| Talent marketplaces | Access to individuals and the contracting rails | €300–600 | Someone leaves in week five and nobody backfills |
Rates move with role and seniority as much as with category. Our current day rates by role are published rather than quoted on request.
Why the spread is that wide
Pyramid staffing explains most of it. A large integrator or consulting practice runs one senior to four or six juniors, bills you a blended rate, and gives you output weighted towards the juniors. That is efficient on a workload of 200 similar tickets and expensive on a workload of twelve hard ones.
Bid cost is the second mechanism. Winning a seven-figure managed-scope contract involves months of pre-sales, solution architects and legal review. That cost is recovered inside the rate card of the contracts that are won, which is why the same engineer costs 60% to 100% more through a large firm than through a provider that sells by the day.
Third, fixed price carries contingency. When a supplier commits to a scope it cannot fully see, it prices the uncertainty, typically 20% to 40% on top of its own estimate. You pay that premium whether or not the risk materialises. How each category of AI consulting firm prices the work goes further into where the overrun lands.
What breaks when the work is AI or data
Four failure modes recur, and they are category-specific rather than firm-specific.
- Pyramid staffing assumes work decomposes into junior-executable tickets. Retrieval quality, feature pipelines and evaluation harnesses do not decompose that way until well after the first production release.
- Fixed scope assumes requirements hold. They rarely survive the first evaluation run against real data, and each change request costs two to four weeks of contract administration before anyone writes code.
- Offshore delivery centres with a five to nine hour time gap turn a 20-minute question into a day of latency. On exploratory work with twenty questions a week, that is the whole schedule.
- Data residency is a contractual question, not a technical one. If personal data cannot leave the EU, the supplier’s delivery location decides whether you can use them at all, before any skills discussion.
How to choose in one pass
- Scope is clear, volume is high, duration is years: global systems integrator.
- You need a regulator or a board to accept the answer: a consulting practice, for the opinion rather than the code.
- You know what to build, you lack the people, and your own engineers will review the work: staff augmentation from a nearshore provider.
- You need commodity run support around the clock: offshore delivery centre.
- You need one specialist for three weeks: a marketplace.
The nearshore versus offshore trade-off is mostly about overlap hours and legal exposure, not about rate.
Where renting engineers is the wrong answer
If you have no internal technical owner, do not rent engineers. Time-and-materials teams need someone on your side to set priorities and accept work; without that, you will pay for motion. If you need a supplier to carry delivery risk on a fixed budget, we are the wrong shape, and you should pay the contingency premium to someone who will sign for it. If the task is under four weeks, the ramp-up cost swamps the saving. And if the requirement is 24/7 run support rather than engineering, a delivery centre will do it better and for less.