Comparisons · · 3 min read

Four ways to buy AI engineering capacity, compared

The four sourcing models for AI engineering compared on cost, speed, accountability and exit risk, including the cases where staff augmentation is the wrong answer.


Most buyers compare vendors. The more useful comparison is between sourcing models, because the model sets the ceiling on what any vendor inside it can do for you. There are four in common use, and they fail in different ways.

The four models

Global systems integrators and Big-4 practices. You buy a statement of work, a partner, and a delivery methodology. Day rates typically run two to four times the European nearshore band. You are paying for indemnity, procurement familiarity and the ability to point at a large balance sheet if the programme fails.

Boutique AI consultancies. Fixed-scope projects, senior-weighted teams, strong opinions about architecture. Rates sit between integrators and augmentation. They are good at defining the problem and poor at staying once it is defined.

Talent marketplaces. You get a vetted individual within days at close to the contractor’s own rate. There is no team, no shared context, and no one accountable for the outcome above the individual.

Staff augmentation and nearshore teams. Engineers join your team, your standups and your repo. You keep architectural control. The vendor carries recruitment, employment, compliance and bench risk.

Where each one actually wins

Your constraintThe model that fits
You don’t yet know what to buildBoutique consultancy
You know what to build, and lack handsStaff augmentation
One specialist, one narrow gap, this weekMarketplace
Board-level risk transfer and procurement coverSystems integrator
Regulated data that cannot leave the EUEU-based augmentation or in-house

Read that table as a diagnostic rather than a ranking. The most expensive mistakes we see are buyers who use a consultancy to do augmentation work, and buyers who use augmentation to do discovery.

The four questions that decide it

Ask these before you shortlist anyone:

The cost comparison people skip

Day rate is the visible number and rarely the decisive one. The real comparison has four lines:

  1. The rate itself. See the daily rate benchmark for the current European bands by role and market.
  2. Management overhead. Every engineer you add consumes senior time. Consultancies absorb some of it and charge for the privilege. Marketplaces absorb none.
  3. Ramp time. Weeks of partial productivity before an engineer contributes. Longer for anyone joining a fixed-scope team that will hand over later.
  4. Rework. The cost of the model getting the problem wrong, which correlates with how far the decision-maker sits from the code.

A boutique consultancy at twice the day rate can still be cheaper on a three-month discovery engagement. An augmentation team at half the rate can be more expensive if you have nobody senior to direct it.

Where staff augmentation is the wrong answer

We sell augmentation, so treat this section as the honest part.

What to do next

Write down your answers to the four questions above before you contact anyone. Buyers who arrive with those answers get a shortlist of two. Buyers who arrive asking “what do you charge” get a proposal from every model at once and no way to compare them.

If augmentation is where you land, the how it works page covers the engagement mechanics, and nearshore vs offshore covers the geography question that usually follows.