Comparison

Agency or in-house?

We are an agency, so read this accordingly. It includes the cases where hiring beats engaging us, because those cases are real and pretending otherwise wastes everyone's time.

THE COMPARISON

	Hire in-house	Engage an agency	Fractional
Time to start productive work	3–6 months	1–2 weeks	1–2 weeks
Annual cost (senior, EU)	€120–180k loaded	Per project	€4–9k per month
Domain knowledge accrual	Deep, permanent	Partial, transferred	Good, ongoing
Breadth of experience	One person's history	Many projects	Many projects
Availability	Full time	Project duration	Agreed days
Risk if they leave	Significant	Contractual handover	Contractual handover
Can you evaluate them?	Hard without AI expertise	Judge by delivered work	Judge by delivered work
Scales down	—	✓	✓

Illustrative ranges, not quotes. The salary figure is our own unsourced estimate for senior AI engineering in Western Europe as of 2026, loaded to include employer contributions, equipment and recruitment; check it against current market data for your country. The fractional figure is the published range on our pricing page.

WHEN TO HIRE INSTEAD

AI is your product

If the AI capability is the thing customers pay for, that expertise belongs inside the company permanently. Outsourcing your core competence is a strategic error regardless of short-term convenience.

The work is genuinely continuous

Past roughly full-time ongoing demand, an employee is cheaper and accumulates context an external party cannot match. We will tell you when you have crossed that line rather than quietly billing through it.

Deep domain knowledge is the bottleneck

Where the hard part is understanding your industry rather than the AI, someone embedded permanently will outperform a specialist who has to learn it each engagement.

You can attract and assess talent

If you have a strong technical brand and someone able to evaluate AI engineers, hiring is a solved problem and the calculus favours it. The difficulty for most companies is the assessment, not the salary.

WHEN AN AGENCY MAKES MORE SENSE

You need it working before you could finish hiring. A senior AI search commonly takes three to six months from opening the role to productive output. If the deadline is sooner, hiring is not an option on the timeline, whatever its long-run merits.

The project is bounded. Building one system does not justify a permanent role. Hiring for a project and then having no work is worse for the person than for the company.

You cannot yet evaluate candidates. Without someone who has built AI systems, interviewing for the skill is close to guesswork. Engaging first — and having your team work alongside — makes the eventual hire much better informed.

You want the decisions made once, correctly. Architecture, evaluation strategy and model selection are decided early and are expensive to reverse. Buying experienced judgement for that window is high leverage even if everything afterwards is in-house.

The pattern we see work best: engage for the first system, pair your engineers alongside, and hire once the ongoing workload is clearly full-time. You get speed now and capability permanently, and the hire arrives into a codebase your team already understands.

QUESTIONS

QUESTIONS — 4

Higher day rate, far lower total commitment. On our own unsourced 2026 estimate a senior AI engineer in Western Europe runs €120–180k fully loaded, plus three to six months of unproductive search. A bounded engagement runs from around €8–25k for a prototype sprint to €30–120k for a production build — the ranges published on our pricing page — and delivers in weeks. Past sustained full-time demand the employee is clearly cheaper, and that crossover is the real decision point rather than the day rate comparison.

Insist on it contractually and check it is happening: code in your repositories from day one, infrastructure as code in your accounts, documented architectural decisions, and your engineers pairing on the work. Any agency resisting those terms is optimising for lock-in — that is a reason to walk, and it applies to us as much as anyone.

That is usually the strongest arrangement. Agency for the initial architecture and the first system; your team learning alongside; a permanent hire when the workload justifies it, arriving into something already built and documented. It costs slightly more than either alone and removes the biggest risk in both.

Good — that is the intended trajectory. We fold them in, transfer context deliberately, and scale our involvement down as they ramp. An agency that treats a client hire as a threat has the wrong business model.

Keep reading

30-minute call · no deck