The Question Behind the Question
At some point, every managing director who takes AI seriously asks the make-or-buy question: build an in-house team or buy externally? The honest answer starts with a counter-question: is data analysis your business — or a tool for your business?
For a machine builder who wants to improve production, it is a tool. And for tools, an old rule applies: you buy the capability, not the factory that produces it.
What an In-House Team Actually Costs
The calculation is routinely underestimated because it stops at salary. In full, it looks like this:
- People: an experienced data scientist or ML engineer costs a high six-figure annual budget over time — and one alone is not enough, because a lone fighter without a sparring partner is a cluster risk (vacation, resignation, blind spots).
- Recruiting: these profiles are contested; six to twelve months typically pass between decision and productive employee. Mid-sized manufacturers compete against corporate and tech salaries for them.
- Ramp-up: even an excellent data scientist knows neither your machines nor your processes on day one. Another half year typically passes before the first productive result.
- Infrastructure and operations: hardware, toolchain, model maintenance, cover arrangements — the invisible second half of the cost.
Realistically, we are talking about a mid six-figure annual budget and 12–18 months to the first reliable result. For companies with permanently high analysis demand, that can pay off. For a first concrete problem, it is the most expensive conceivable solution.
What Stands Against It
For comparison, the fixed-price path: a bounded pilot at a fixed price in four phases with one of the three applications — for instance Process AI for predictive maintenance. Result: a trained model for your equipment that you own, an accuracy report, a readiness assessment. What drives the prices is laid out openly here.
The difference is not just the amount — it is the risk distribution: price and duration are fixed before the start, and whether more is worth it is a decision you make after the first result, not after the first year.
The Three Decision Questions
- How many analysis projects do you have per year? One or two → commission a pilot. A dozen, across all areas, permanently → think about a team.
- Is the knowledge strategically exclusive? If the analytics itself is meant to become your competitive advantage (not just its results), that argues for internal build-up — long term.
- Can you retain and utilize the team? An AI team without a steady supply of problems walks away — and then the calculation starts over.
The Underrated Third Option
Make or buy is really not an either-or, because on the fixed-price path the result belongs to you: models, reports, data preparation. If you later build your own team, it inherits a working foundation instead of a green field — your first own data scientist starts with a running system and an accuracy report instead of an empty repository.
Buy first, make later — on the basis of evidence instead of hope. That is the order we recommend even when your long-term goal is your own team.
The cheapest starting point for both paths is the same: a bounded pilot shows whether and where analytics pays off for you at all — open-ended. Talk to us.