AI assistant

The AI assistant is a chat panel that answers questions about your organization's own data, projects, quotes, customers, products, tickets, tasks, plans, and offers, in plain language, and can propose creating a customer, task, ticket, or RMA for you to approve. It is not a general chatbot: everything it answers comes from your real records, not the open web.

Before you start

The assistant only appears if an owner has switched it on for your organization, in Settings → Preferences → AI assistant. If it's off, the chat panel does not render at all and the underlying route refuses requests, so no data reaches the AI provider either way.

Ask a question

Click the chat panel (bottom-right) and type a question, such as "what's the budget on the Meridian project" or "how many open tickets does this customer have." Simple questions are answered instantly from your data with no AI call involved; anything the built-in rules can't confidently handle falls through to the AI, which reads your records through the same permissions your account already has, then writes a plain-language answer.

What it can and cannot see

• It only reads data your role already has view access to. It cannot see more than you can see in the app yourself.
• It has no tool that reads device credentials or API keys, under any circumstance. Those are excluded by design, not just by permission.
• It is bounded to a fixed set of read tools (projects, quotes, customers, products, tickets, tasks, plans, notes). Questions outside that scope get "can't answer" rather than a guess.

Creating or changing something

Ask it to create a customer, task, ticket, or RMA, or to edit a field on an existing one, and it prepares a proposal card showing exactly what it intends to do. Nothing is written until you click Approve. Discard cancels it with no changes made. Deleting a task or ticket works the same way, with an extra destructive confirmation, and is off by default until an owner turns it on.

Two more organization-level switches sit on top of your own role's permissions: whether the assistant may propose edits at all, and whether it may propose deletions. Both are set by an owner in Settings → Preferences, and the assistant can never exceed either the organization's switches or your own role's edit permission, whichever is stricter.

Tips

• Each question and answer respects your daily usage limit; hitting it returns a clear message rather than failing silently.
• Approving a proposal is a normal write to your data, gated the same way as doing it by hand, it is not an unlimited or hidden channel for changes.
• Turning the assistant off in Settings removes it for your entire organization immediately, not just for you.