BrainMotor guide
Independent AI on your company’s server
Understand what your company controls with self-hosted AI, from source code and backups to model choice.
Identify what remains under your control
An on-premises server and a VPS in your company’s name are two ways to host AI under your control. In both cases, establish who administers the machine, holds the keys and restores backups. With BrainMotor, the instance’s code and data are installed on that infrastructure.
During the pilot, verify that you can retrieve the configuration, source code and a usable backup. Document how to revoke maintenance access. This lets you change service providers while keeping your instance and the installed version.
Choose where each request is processed
Local storage and local processing are separate choices. A model installed on your server can process a document there when its tools and sources stay local. When you use an external API, the data needed for the request goes to the provider you chose.
Before enabling an external connector or model, record the fields sent, their destination and the rules that apply. Compare models on the same task to assess answer quality, latency and cost. Different models can produce different results.
Test independence on a concrete task
Start with a small corpus, such as approved internal procedures. Ask the agent to find the applicable procedure and cite its source. Try a document it should not be able to access and verify that access is denied. Then test restoration and, if the project calls for it, switching models.
The pilot helps decide which flows, models and permissions to authorise before expanding the scope. Integration compatibility is checked in your environment.