Platform · Choosing a provider

Swiss AI providers.
Six questions before you choose.

Almost every provider in the Swiss market advertises sovereignty. These six questions separate the ones where the word holds from the ones where it stays marketing.

Request early access Book a call
Starting point

Sovereignty has become the standard promise.

The Swiss market for AI platforms has grown quickly, and practically every provider now leads with data protection, Swiss hosting and compliance. That makes choosing harder rather than easier: when everyone promises the same thing, the promise stops carrying information.

So we do not compare marketing claims but six properties that can be verified and that make a real difference in operation. They work as a checklist for any conversation — with us as much as with anyone else.

The criteria

Six questions worth asking every provider.

Each one comes with why it matters and how AIgent answers it. We deliberately do not judge how the other providers answer — those answers are best collected from them directly.

Who owns the company running the servers?

Does the platform run on its own infrastructure, or on a US hyperscaler?

A Swiss region of Azure, AWS or GCP sits in Switzerland but belongs to a US company, which brings it within reach of the US CLOUD Act. The location of the data centre alone says nothing about the legal position.

At AIgent: AIgent runs in Swiss data centres or on your own infrastructure, independent of US hyperscalers.

Is on-premise real, and from how many users?

Is on-premise an actual option, or an enterprise promise that starts at a four-digit seat count?

On-premise appears on many datasheets. What matters is the size at which it is genuinely offered, and whether it goes as far as air-gapped when banking secrecy or health data demand it.

At AIgent: AIgent runs on your own Kubernetes in house, up to air-gapped, with no minimum size set by enterprise sales.

Can you choose the model freely?

Is the platform tied to one model vendor, and can the model be set per use case?

Model lock-in decides where your most sensitive data goes. A provider offering a single model moves that decision away from your compliance function.

At AIgent: AIgent is model-independent, Swiss open-source models included, and the model is chosen per agent according to data class.

What does the price include?

Which AI usage is included, and which services are priced separately?

Compare seat types, included usage and separately priced services. Reliable planning needs a clear usage allowance and clear costs for additional usage before signing a contract.

At AIgent: AIgent Cloud offers Standard/Plus seats with an included usage allowance. On-premise has a fixed organisation licence; implementation and infrastructure are priced separately. Contact us to discuss pricing and usage.

Is a company your size even the intended customer?

Is the offering aimed at large corporates, at a single industry, or also at SMEs and the mid-market?

Part of the Swiss field is built for one industry or for large enterprises. That is no weakness on their side, but it decides whether you as an SME will ever see a contract that fits.

At AIgent: AIgent is horizontal by design and built for SMEs through mid-market, not only for thousand-seat deals.

Is the output reproducible?

Is a recurring task handled by tested code, or decided afresh by the model on every run?

For answering questions, a language model is enough. For a process that has to withstand an audit you need the same result for the same input, and a trail showing which model processed what.

At AIgent: AIgent runs skills as tested code in an isolated sandbox, with every step visible in the run history.

Honestly placed

No single criterion decides it.

Several Swiss and European providers meet individual points as well as we do, some better within their niche. There are excellent providers with real on-premise, and others specialise in one industry and go deeper there than any horizontal platform can. Anyone claiming to have no competition has not looked at the market.

Our difference is the combination: independent of US hyperscalers, real on-premise on our own infrastructure, free model choice per agent, clearly separated cloud and on-premise pricing models, open to SMEs and the mid-market rather than only large deals, and reproducible results through tested skills. Individual points you will find elsewhere too. All six together is where the case for us is made.

Questions & answers

Common questions on choosing a provider.

Which AI providers are there in Switzerland?
The Swiss field has grown considerably and falls roughly into four groups: infrastructure and model providers supplying compute or language models; industry specialists, for instance in finance or law; secure assistants for highly regulated environments; and horizontal platforms on which companies run their own AI agents. The field includes Alpine AI, Unique, DeepJudge, Infomaniak and Swisscom, alongside AIgent. This page deliberately does not rate these providers, because feature sets and pricing change constantly. It gives you the six questions to assess any provider yourself, with current answers.
Is a Swiss Azure or AWS region already sovereign?
Not entirely. The data sits physically in Switzerland, but the operator is a US company and therefore subject to the US CLOUD Act, which under certain conditions gives US authorities access to data in the custody of US firms regardless of where it is stored. For many use cases a Swiss hyperscaler region is still a sound choice. Where data control has to be legally watertight, however, what counts is not only the location but who owns the operator.
Is Apertus a Swiss AI provider?
Apertus is an open language model developed in Switzerland, not a platform provider. You cannot buy it as a finished solution: you need a platform that runs it, connects it to your data, assigns permissions and logs its use. That is exactly the role AIgent fills — the model is freely selectable, Swiss open-source models included, and is set per agent according to data class.
What does sovereign AI actually mean?
Sovereign means you decide where and under which legal framework processing happens, and can evidence that decision. In practice that is four things: data is processed in Switzerland or on your premises; the operator is subject to no foreign disclosure claim; you determine per use case which model sees which data; and every step is logged. If one of the four is missing, sovereignty stays a marketing word.
What should regulated industries watch for?
Banks, insurers, healthcare and public administration work with data whose processing is restricted by the revised Data Protection Act, supervisory law or professional secrecy. For them three of the six criteria are not negotiable: the independence of the operator, real on-premise up to air-gapped for the most sensitive data, and a complete audit trail. Choosing the model per agent then lets sensitive cases run in house while everyday tasks use more capable models.
Swiss cloud or on-premise?
That depends on the data class, not on principle. For most tasks, hosting in a Swiss data centre delivers sovereignty without you operating infrastructure yourself. For particularly sensitive data AIgent runs fully on-premise, up to air-gapped. Because the model is chosen per agent, both can be mixed within the same company instead of deciding once for everything.
On-premise in detail Sovereignty explained Start the AI check

Sovereign AI. For every discerning company.

Request early access for your business. Already planning a specific AI project? Book a call to discuss your use case, data sources and the right deployment: Swiss cloud or on-premise.

Request early access Book a project call

Invitation by email when your access is ready. You can book a project call now.