What it powers
The admin assistant reads a plain-language instruction typed into the command palette, works out a plan against the platform’s own data model, shows that plan before doing anything, and only executes after you confirm. Every action stays inside your permissions and lands in the audit log.
Why the model is swappable
The key and the model name live in one place in the configuration. If your organization already standardizes on Anthropic for other tools, or simply prefers Claude’s behavior for this kind of task, switching is a settings change, not a deployment.
Claude is one of four options on the provider list, alongside GPT, Gemini and DeepSeek. All four run the same assistant against the same instruction typed into the command palette; what differs is the provider and the model actually answering, not the scope of what the assistant can do in the panel.