Definition
A deployment mode where the software and data run on the company's servers, on its premises or in its datacenter, rather than in a provider's cloud.
An on-premise deployment installs the AI platform directly on the company's infrastructure: its physical servers, its datacenter or its private cloud. Prompts, documents and search indexes never pass through a third party, it is the strongest confidentiality guarantee that exists.
Long reserved for large groups for reasons of complexity, on-premise has become widely accessible: a platform like Helix installs in a day on standard hardware, with open-source models running locally and, if desired, controlled access to cloud models for non-sensitive tasks.
A laboratory installs a server fitted with a graphics card in its own server room. The AI assistant runs there, reachable only from the internal network.
When data must not leave the building, when the site has little or no connection, or when you prefer a one-off investment to a subscription.
Do you need an IT team for on-premise?
Not necessarily. Installation and maintenance can be entrusted to a service provider, working remotely or on site. What you mainly need is a suitable room and an internal point of contact.
All AI models, deployed on your infrastructure, with your documents connected.