At a glance
| Criterion | In-house team | External AI partner |
|---|---|---|
| Time to first result | Months, once recruitment and onboarding are done | About 4 weeks for a pilot on your real data |
| Cost structure | Salaries, recruitment, training and tools, also between projects | Fixed project prices, for example a pilot from CHF 4,900, plus optional Care from CHF 290 per month |
| Knowledge of your business | Deep, built up every day | Must be learned, needs your input and time |
| Experience with AI projects | Grows with each project you do | Brings experience from comparable projects |
| Availability | Holidays, sick leave and departures hit a small team hard | Agreed by contract, but shared with other clients |
| Long-term control | Full control over code, data and priorities | Depends on documentation, access rights and a clear handover |
| Best suited for | AI as a core part of your product or a steady stream of projects | A few defined use cases and fast, measurable results |
The strengths of an in-house team
Your own people know your customers, your processes and your exceptions. They are in the next office when something breaks, and what they learn stays in the company. If AI is part of what you sell, or if you already see a long list of projects ahead, building your own capability is often the right long-term decision.
What building in-house really takes
- At least one person with experience in language models, data, integrations and security, which is rarely a single profile.
- Time to find and hire them; experienced specialists are in demand and not easy to recruit in Switzerland.
- A plan for when that one person is on holiday or leaves, because a team of one is a single point of failure.
- Hosting, monitoring, access control and an audit trail, not only the AI itself.
- Time to keep up with models and tools that change every few months.
The strengths of an external partner
A partner can start right away and brings experience from similar projects, including the mistakes already made elsewhere. You pay for a defined result instead of a permanent position, and with fixed prices you know the cost before you start. A good partner also tells you when a project is not worth doing.
The downsides of a partner
- A partner has to learn your processes, and that takes time from your team.
- There is a risk of depending on one provider if the work is poorly documented.
- Knowledge can leave with the partner if nobody internal is involved.
- A partner works for several clients, so response times must be agreed in writing.
These risks can be reduced. Insist on documentation, training for your team, access to code and accounts, and a clear handover plan. At Linxa, the Production package includes team training and documentation.
The combination that works for many SMEs
In practice, many small and medium companies do best with a mix. One internal owner, who does not need to be technical, knows the processes, sets priorities and approves results. The partner builds, integrates and runs the solution. Over time, your team can take over more of the work, or keep the partner for operation only.
Our verdict
Build in-house when AI is central to your business, when you have a steady pipeline of projects and when you can hire and keep the right people. Work with a partner when you have a few clear use cases, need results within weeks and have no AI specialist on staff. If you want to build your own capability step by step, combine both: a partner for the first projects and a clear plan to hand over knowledge.