An Orkeste agent reads the newest Zendesk tickets every hour, works out the topic, the urgency and the queue each one belongs to, and writes that in an internal note on the ticket. Your team still assigns the ticket; the agent saves them the reading.
The problem
Someone starts every morning by opening each new ticket to decide what it is about, how urgent it is and who should take it. Zendesk triggers handle the obvious cases on keywords, but customers rarely describe a problem with the words your rules expect.
How the workflow runs
Runs: Every hour, or when you start it.
Find new tickets: The agent searches Zendesk for tickets created since its last run, up to five per search.
Read each ticket: It reads the subject and the latest messages of the thread.
Classify: It decides the topic, the urgency and the queue, using the categories written in its briefing.
Write an internal note: It adds an internal note with the category, the priority, the suggested queue and a one-line summary.
Flag urgent tickets: Tickets that match your urgency rules are posted to the support channel in Slack.
A person assigns: Your team reads the suggestion and assigns the ticket in Zendesk.
The agent only adds internal notes. Assignment stays with your team, and you can also hold each note for approval in Validations.
Who does what
Triage agent: Reads new tickets and writes its suggestion as an internal note, which customers never see.
Support lead (human): Assigns the ticket and corrects the briefing when a suggestion is wrong.
What this workflow does not do
Does not assign, tag, merge or close tickets.
Reads up to five tickets per search and the latest messages of each thread.
Runs on a schedule or when started, not the instant a ticket arrives.
Can the agent assign tickets to a group or a person?
No. Agents add public replies and internal notes in Zendesk; they do not change the assignee, group, tags or status. The note tells your team where the ticket should go.
Does it replace Zendesk triggers?
No. Keep triggers for hard rules such as routing by channel or language. The agent covers the tickets your rules cannot classify because it depends on what the customer actually wrote.
How many tickets does it handle per run?
Each search returns up to five tickets. On a busy queue, run the agent every hour, or narrow the search in its briefing, for example to one brand or one form.
How do I define the categories?
In the briefing, in plain language: the list of queues, what belongs in each one and what counts as urgent.