AI assistant

• find contacts, conversations, deals and tasks and sum up what is going on with them;
• reply to a customer in a messenger, create or move a deal, change a contact, set a task;
• build or change a funnel, prepare a broadcast;
• work through your file (XLSX, CSV, PDF): count, compare, import contacts.

The assistant runs on Claude models from Anthropic.

Assistant in the menu opens the page with your dialogues, and the assistant button on every other page opens it in a window. The assistant knows which page you are on. New dialogue starts a conversation, with a choice of model: Fast, Standard or Complex task.

Before the first message, every staff member reads and accepts the terms of working with the assistant.

The assistant acts with your permissions: whatever you cannot reach, it cannot either. It reads without asking, but before every change it shows a card with exactly what will change. Approve runs it; Decline, with a note on what is wrong, makes the assistant suggest something else.

Auto-approval in a dialogue lets the assistant run every action without asking. You approve them in advance and are responsible for them; limits and role permissions still apply.

Drag a file into the message box, paste it with Ctrl+V or pick it, up to 30 MB. A large file does not pass through the model whole: the assistant processes it in its isolated sandbox. This is how you import contacts from Excel or another CRM's export: the assistant first shows how many are new, matched or invalid.

An administrator sets limits per month for the business and per dialogue; the assistant never spends more. You can close the tab: the work goes on and the result waits in the dialogue. A dialogue with no messages for two weeks closes by itself.

What the assistant reads in the CRM, and your files, are processed by Anthropic (USA). Never send the assistant passwords, access keys or other secrets. For billing, see AI credits.