SleekAI's chatbot lives on WordPress and reads your real data: wp_posts, wp_postmeta, taxonomies, ACF fields, and any custom tables you map. When the visitor asks for a human or the model flags low confidence, an action hook calls the Intercom Conversations API and pushes a new conversation with the transcript, the contact record, custom attributes like plan tier or role, and the originating page URL.

The handoff creates or matches an Intercom contact by email, so the conversation joins that customer's existing thread history rather than spawning a new identity. Custom attributes you map from WordPress flow into the contact record on creation, which means assignment rules, Series, and reporting in Intercom see the same shape they would for a chat started in Messenger.

Display conditions decide which bots can escalate. Marketing-page bots stay deflection-only. Refund and billing bots open Intercom conversations routed to a specific team inbox with tags attached. SleekAI's own conversation log captures every exchange with model, tokens, and origin page so you can tune deflection without leaving WordPress.

Intercom is built around the contact record. Plan tier, signup date, last seen, custom attributes, and conversation history all hang off it, and the whole Operator, Series, and assignment machinery reads those fields constantly. A chatbot handoff that just emails a support alias bypasses all of that.

The contact is unidentified, the assignment goes to round-robin, the Series never fires, and the agent picks up a case with no history attached. SleekAI flips that path by calling the Conversations API directly, matching or creating the contact by email, and updating the WordPress-side custom attributes onto the Intercom record before the conversation is created. The team_assignee_id picks the right inbox.

Tags fire the right view. Custom attributes populate the right Series. The agent opens the conversation and sees the same shape they would for a Messenger inbound, with the chat history already threaded as the first part.

That structural fidelity is what makes deflection actually scale, because the cases the bot does not solve still arrive in the right place with the right context, which is what every Intercom team wishes their chatbot already did.