Knowledge assistants
Research, consolidate and make company knowledge available.
Swiss AI Studio: the control centre for company-wide AI.
Studio connects models, agents, company knowledge, applications and compute infrastructure into one controllable environment. Here you define which AI handles which task and which knowledge it may access. You also set which tools it can use and when a person has to approve.
In brief: Studio answers the question “Who may do what, with which model, on which data – and who approves?”. Studio, Agents and parts of Factory run in our own development and operating environment and are continuously extended. Further operating models, multiple nodes and private and local installations will follow step by step. Which building blocks suit your business depends on the use case, infrastructure, data protection and interfaces. Publicly, we show the operating principles, not security-critical details.
Starting point
From individual AI tools to connected infrastructure.
Companies gradually adopt different models, assistants, automations and data sources. Without shared control, you get isolated solutions, inconsistent permissions and unnecessary costs. You also get workflows that nobody can trace afterwards.
Studio adds a shared technical layer underneath: models, agents, knowledge, tools, infrastructure and governance are not run side by side but connected. That is why Studio is neither another chat nor a collection of automations. It is the control layer of the Swiss AI Platform.
Models
The AI stays interchangeable. The platform stays.
Studio decouples the business process from any single model. It defines which model is intended for which process and under which conditions it may be used. This way, the architecture does not depend on one provider, one model or one leap in technology.
- Model Routingthe right model is assigned to a task or an agent.
- Local and private modelssensitive tasks can run on your own or dedicated infrastructure.
- External modelspowerful services used specifically where they bring an advantage.
- Controlled switchingModels can be tested, replaced or reassigned without rebuilding the application.
Agents
A language model becomes a unit of work.
In Studio, an agent gets a clear framework: task, knowledge, model, tools, rules, approval, result. It can research, analyse documents, fetch data from an ERP, prepare a case or trigger a defined automation.
The other half is what matters: what an agent may not do is defined just as clearly as what it may do. Actions outside the verified scope are to be blocked and logged.
Insights
Company knowledge becomes part of the infrastructure.
AI only becomes useful once it knows the context of your business. Studio connects agents and models with the sources approved for each task – by role and context, not across the board.
- Documentation and manuals
- Contracts and business documents
- Product information
- Project knowledge
- CRM and ERP data
- Internal policies
- Structured databases
- Knowledge bases and RAG systems
Applications
One Studio, many AI systems.
They all run on the same controlled infrastructure. On top of that come custom applications for individual processes, industries or business units.
Document AI
Recognise, classify, analyse and process documents.
Process agents
Monitor and prepare workflows and carry out defined steps.
Research and analysis
Prepare information from internal and external sources in a structured way.
Customers and communication
Handle enquiries, provide information, prepare replies.
Management intelligence
Condense data and developments for decision-making.
Orchestration
Several building blocks, one traceable workflow.
A single model rarely solves complex tasks. Studio lets several specialised building blocks work together – individual functions become controlled workflows.
Trigger
An email, a document or a schedule starts the process.
Insights
The agent fetches the information approved for its task.
Model and tools
Studio assigns the right model and grants access to the permitted systems.
Checking
A second agent checks the result before it moves on.
Approval
Where necessary, a person decides – only then is the action triggered.
Handover and log
The result goes into the target system, and the whole process stays documented.
Tools
Each agent gets only what it needs.
Not every AI gets access to every system. Studio defines which agent, which tool, for which task, with which rights.
- ERP and CRMMaster data, customers, cases
- DMS and drivesDocuments and filing
- Email and calendarInbox, sending, appointments
- TelephonyCalls and callbacks
- Databasesstructured records
- Web and interfacesexternal sources and systems
Human in the loop
Automate without giving up control.
Not every process should run fully on its own. Studio allows defined control and approval points: an agent analyses and prepares, and the final action only happens once a responsible person approves it.
This applies especially to payments, customer communication, contract and document processes, critical data changes and actions in third-party systems. The AI does the work, people decide.
Governance
What is defined in Studio.
Professional AI does not just mean powerful models but clear responsibilities. This way, governance is not added afterwards but is part of the architecture.
| Question | Defined as |
|---|---|
| Who may use which agent? | Role and permission |
| Which model may process which data? | Data class and model rule |
| Which tools may an agent use? | Tool approval per agent |
| Which actions need approval? | Approval point in the workflow |
| Which processes must be logged? | Logging rule |
Operation
Know what the AI is doing right now.
The more models and agents are in use, the more important an overview becomes. What is visible depends on the specific implementation.
- Active agents and running processes
- Pending and cancelled tasks
- Models used per task
- Run times and response times
- Model and infrastructure costs
- Compute node utilisation
- Results and measured quality
- Approvals and human interventions
Operating models
From cloud to on-premise.
The architecture can be adapted to data protection, speed, cost and the needs of your business. Additional models, agents, knowledge sources and compute nodes can be added without building separate infrastructure for every new use case.
Cloud
Use external AI and infrastructure services.
Private AI
Dedicated environment for higher requirements for control and data protection.
View detailsOn-premise
AI infrastructure on your own premises.
View detailsHybrid
Local models combined with selectively used external services.
View details
Swiss AI Platform
Studio controls, Agents work, Factory provides.
Studio is part of the Swiss AI Platform. Together, they form end-to-end infrastructure – from computing power through the agents to central control.
Swiss AI Studio
Control and governance: configuration, orchestration, models, knowledge, permissions, operation and approvals.
Swiss AI Agents
Execution: specialised units of work that analyse tasks and carry out defined workflows.
View detailsSwiss AI Factory
Computing power and runtime: models, compute nodes and the technical foundation for operation.
View details
Suitability
Where Studio pays off.
Studio pays off where several AI applications are being built and need to fit together. That means when different models are in use, internal knowledge has to be integrated securely, agents access company systems and workflows are automated.
It also pays off where human approvals are needed and private or hybrid operating models play a role. The same applies where the environment must stay controllable and expandable for years. For a single, self-contained application, the effort is too great – a product or a standalone solution is the better route.
Frequently asked questions
Frequently asked questions about Swiss AI Studio
Is Studio a separate product?
Studio is part of the Swiss AI Platform and runs with our products and custom solutions. Whether and how clients work directly in Studio depends on the project.
Can I see what the AI has done?
Yes, that is the purpose of the operations view: tasks, results, model used, costs – all traceable.
What is the difference between Studio, Agents and Factory?
Studio controls and monitors, Agents execute tasks, Factory provides models and computing power. Three layers, three responsibilities.
Which components do we need?
That depends on the use case, existing infrastructure, data protection and the interfaces required. This question comes at the start of architecture consulting, not at the end.
Do you show internal details?
Publicly, we show the principle, layers and data flow at a high level. Internal topology, routing rules, security rules and version states stay internal.
Is it already in production?
We state the maturity of each building block openly. Phrases such as “is designed to” describe the architecture, not a client deployment. Studio, Agents and parts of Factory run in our own development and operating environment and are continuously extended. Further operating models, multiple nodes and private and local installations will follow step by step.
Next step
Your AI infrastructure starts with the architecture.
We do not start with a single model or a single agent. We look at the process, the data, your existing systems and the requirements for control and operation. From this, we build an architecture that fits your business.