AI that acts. Data that stays.

Botster is an agentic AI platform from Switzerland and a product of Caprilytics AG. It connects approved company knowledge with tools and business systems. Botster delivers answers with sources and can carry out permitted tasks. Roles, autonomy limits, approvals and an audit trail determine what the platform may know and do.

A classic chatbot mainly writes answers from its model context. Botster can additionally search approved company knowledge, show sources and use controlled tools. That turns a chat into a bounded AI agent for multi-step tasks. The difference is not only the language model, but permissions, approvals, system integration and traceable execution.

Yes. Botster can search approved policies, manuals, contracts and other knowledge sources and deliver answers with concrete evidence. Users see which documents or sections a statement is based on. Access rights are respected, so a user cannot reach content through the AI that they are not entitled to in the source system.

Botster supports IT teams with structured troubleshooting. The platform checks active incidents and announcements, searches Microsoft knowledge and finds similar existing tickets. From that come targeted follow-up questions and the next checks to run. The current Matrix42 integration reads data; creating, changing or closing tickets automatically is not yet a productive part of it.

Botster can carry out permitted tasks within configured limits. Depending on the tool, the risk and the autonomy level, a function stays blocked, is only proposed, needs human approval before execution, or may run automatically as a low-risk routine. That produces controlled autonomy rather than an agent with blanket access to company systems.

Botster connects knowledge sources and business systems through controlled skills and APIs. Areas already covered include company knowledge, document processing, ITSM search and troubleshooting, and defined maintenance and operations flows. Which integration can be used productively is assessed per use case. A general interface capability is not automatically advertised as a finished product integration.

Botster runs on Swiss infrastructure by default. The data location, the models used and the subprocessors involved are documented for the chosen configuration. External models are only enabled in a controlled way when this has been agreed and approved. That keeps it transparent which data is processed for which purpose by which provider.

Botster combines user permissions, tenant separation, risk classes and approvals with an audit trail for AI and tool actions. Sources, decisions, calls and results remain traceable. These capabilities support security, revFADP and EU AI Act requirements. The concrete regulatory classification depends on the area of use, however, and is not replaced by a blanket compliance promise.

A Botster pilot costs CHF 5'000 and runs for 100 days with up to 20 users. It starts with a clear use case, defined data sources, one skill module and measurable success criteria. Permissions and autonomy limits are then set up and documented test cases are worked through. The closing evaluation shows which criteria were met and which were missed, which integrations are still open, and the recommendation to stop, adjust or expand. If you sign up for a package within 30 days, the first two months are on us; after that at least ten paid months follow.