How much does it cost to set up private AI?

Describe a piece of work you already do: drafting internal notes, helping with code, or summarising a document you are allowed to use. We can then agree an example that demonstrates whether the chosen setup is useful. A model that launches successfully still needs to be checked against your actual task.

A focused setup should identify the model, where it runs, how you start and stop it, and how you repeat the agreed example. A shared interface, document search, account permissions or connections to business software add work and need their own scope.

Ollama supports both local and cloud models. Its documentation says local prompts stay local, while cloud models send prompts for remote processing. Ollama also provides a setting to disable its cloud features.

The whole setup matters: the chat interface, extensions, document processing, logs and external tools can have their own data paths. Tell us which information must stay on your equipment so the proposed configuration and checks address that requirement.

Driver problems, limited memory, more than one user, a larger document collection and custom software connections can expand the work. Hardware purchases, paid software and any external hosting should be identified separately before you agree to them.

If suitability is uncertain, begin with the hardware assessment. It gives you a way to decide whether to reuse the machine, reduce the task or consider an upgrade before paying for a wider installation.