Ai Copilots FAQ
What AI features exist in the CRM?
A page-aware copilot that answers questions using the record on screen, lead summaries and a lead agent for outreach drafts, meeting-prep briefs, lead parsing from pasted text, WhatsApp conversation analysis, and AI drafts of meeting follow-ups — all powered by Gemini models configured in workspace settings.
What are copilot skills?
Skills are reusable instruction sets (name, description, when-to-use guidance, body, optional allowed tools) stored per workspace and injected into the copilot. Admins manage them under Settings → Copilot; seed skills ship with the system.
What are MCP servers in the integrations registry?
External tool servers the copilot can call. Each entry has an id, label, URL, enable flag, and optional headers (values encrypted at rest). Status probes run a live tools/list check showing connected, unavailable, or disabled plus a tool count.
Does the AI invent data or take actions by itself?
No. Prompts restrict answers to supplied CRM context and say what's missing otherwise; outreach drafting falls back to a template draft if AI output isn't usable JSON; and meeting follow-ups always wait for human review before anything is written.
Who can use which AI feature?
The generic copilot needs settings read; lead summary/agent/meeting-prep only need read access to that specific record; parsing text into new leads requires leads create; skill administration is settings manage and MCP registry is integrations manage.
Where did Threads go?
Threads is now a compatibility redirect — agent runs live in the Agents hub under Runs, with status filters and pagination preserved from old links. Execution history also surfaces in the Work Hub Activity tab.