SleekAI maps fields from wp_posts, wp_postmeta, taxonomies, and any custom table into the chatbot's prompt, so a question about a product, an order, or a documentation page has data to ground on. When deflection runs out, an action hook posts to /api/v2/tickets with the requester email, the subject, the description, the priority, the group_id, and any tags you configure. The transcript lands in the description as plain text, and the page URL ships as a tag plus a custom field if you set one up.

Display conditions in SleekAI scope which bots are allowed to escalate. A blog bot can stay deflection-only, while a billing or refund bot always opens a ticket and routes it to the finance group. Multibot lets product, support, and sales each have their own escalation path with different default priorities and SLA targets baked in.

Every handoff also writes to the SleekAI conversation log inside WordPress with the model used, token usage, originating page, and user role. Pair that with the JS API and PHP embed so a refund flow on the order detail page can trigger a handoff with the order id already filled in, without rebuilding the widget or asking the customer to start over.

Freshdesk runs on dispatcher rules, observer automations, and SLA policies that read structured fields. When a chatbot creates a ticket by emailing a generic alias, none of that machinery fires correctly. The dispatcher cannot route because the group is empty.

The SLA policy cannot attach because the priority is default. The CSM team cannot filter by plan tier because the plan never made it onto the ticket. Agents then triage by hand and the deflection layer ends up costing more than it saves.

SleekAI creates the ticket the way an experienced agent would, with the right group, the right priority, the right custom fields, and a description that already includes the page URL, the role, and the full chat transcript. Once that is in place, the rest of Freshdesk works the way you designed it. Automations route the case to the right team.

Observers fire because the plan tier on the ticket matches the rule. SLA timers attach the moment the ticket is created. And the chatbot stops being a layer the agents work around, because the cases it sends in already arrive looking like the cases they wish all their tickets did.