AI-powered support

🚧 In progress — AI-assisted drafting is live in the agent inbox, and employees can now ground an answer in your knowledge base. Autonomous auto-reply, sent without a human, is not.

What works today

Drafting. In a conversation or a ticket, ask CoreX to draft a reply. It reads the thread and proposes a response you edit and send — a human is always the one who sends. Internal notes can be drafted the same way.

Ticket work from chat. The AI Workforce can list open tickets, read a ticket's detail, and propose creating one from a conversation. The create is confirmation-gated: the employee proposes, you approve, and only then does the ticket exist.

Grounding in your own docs. Employees can search your knowledge base — title and body, with an excerpt returned per match — so a drafted reply is built from what your team actually wrote rather than from the model's recollection. Publish an article and it is searchable immediately; drafts are not.

What's still coming

Confidence-gated auto-reply. Sending without a human above a threshold you set, with a clear hand-off to a person below it. Everything today still ends with a person pressing send.

Why it matters

AI handling the repetitive share of inbound is only safe once it answers from your documented truth rather than from inference — which is why grounding shipped before autonomy.