AI assistant

The AI assistant answers questions about your workspace in plain language — "How many meetings did we book last week, and from which sources?", "Why was this lead disqualified?", "Who on the Sales team has had the fewest meetings lately?" — and explains how to set things up, with links to the right pages.

Open the assistant

Click Ask AI in the top bar of the app, or press ⌘ J (Ctrl J on Windows and Linux). Answers stream in as they're written; when the assistant looks something up, Checked N sources shows which data it used. Your earlier chats are listed on the left, grouped by day and searchable — start a New chat at any time. Copy an answer or ask again with the buttons under it.

Good things to ask:

• Reports — meetings by source, rep or week; no-shows and cancellations; how routers perform.
• Leads and meetings — what happened to a lead, who booked a meeting and through which path.
• Routing — "If [email protected] with 800 employees filled in the demo form, where would she go?" The assistant runs the router as a dry run and shows the path, without saving anything or notifying anyone.
• Fairness — how evenly meetings were shared across a team.
• How-to questions — the assistant searches these docs and links the pages it used.

What it can and can't do

The assistant reads your workspace's data with your permissions: members see what members can see, admins see more. It can't change anything — no edits, bookings, cancellations or deletions. When something should change, it tells you exactly where to do it.

Conversations stay with the person who started them. Each workspace has a daily message limit set by the operator of your installation.

For operators

Platform admins choose the AI behind the assistant under Admin → AI assistant:

1. Add a provider — Anthropic (Claude), OpenAI or any OpenAI-compatible API such as OpenRouter or a self-hosted gateway. Paste the API key and list the models people may use; Test sends a one-line request to check the key and model.
2. Pick the default model, and decide whether people may switch to another enabled model in the chat.
3. Set limits and tone — a daily message limit per workspace, the reasoning effort for Claude models, and extra instructions added to every conversation (for example your support email).
4. Turn it on. The assistant appears for every workspace where the ai_assistant feature flag is enabled.

The page shows how many questions were asked, how many workspaces use it and how many tokens were used in the last 30 days. Keys are encrypted at rest. Messages are sent to the provider you configure, together with the workspace data the assistant looks up to answer.