[ Use case · Support ]
Can Jev route support tickets to the right team?
Yes. Routing a ticket to a queue is a textbook Choice question, and TypeSafe's own docs use this example. In an independent test, 8-way routing cost about 1.5 cents per 1,000 tickets, roughly five times less than small language models.
Is it a fit?
Use Jev when
- You have three or more queues and tickets regularly land in the wrong one.
- Keyword-based routing rules keep breaking.
- You want early warning on angry or at-risk customers.
Skip it when
- You have a single support queue, so there's nothing to route.
- You need a reply sent to the customer. That's a text-generation job.
The questions
One request per ticket, 3 questions. Jev answers them in parallel.
- Choice
Which team should handle this ticket?
Options: Billing · Account access · Bug · How to · Feedback
- Score
How frustrated is the customer?
Scale: Calm → Mildly annoyed → Frustrated → Angry or threatening to leave
- Noul (yes/no)
Is the customer completely blocked from using the product?
Options: Yes · No
Getting started
[ 1 · Test it in the playground ]
No code needed. Paste a real ticket into TypeSafe's playground and ask: “Which team should handle this ticket?” with the options Billing, Account access, Bug, How to, Feedback. You'll see the answer and its probability.
Open the playground (opens in a new tab)[ 2 · Use an existing tool ]
We haven't found a ready-made tool for this workload yet. These directories track what's been built on Jev:
Made with Jev (opens in a new tab)Jev.Store (opens in a new tab)JevHunt (opens in a new tab)
[ 3 · Integrate it ]
To run it on every ticket automatically, call Jev from the system where your tickets live. The developer details below include the full request.
Common questions
- Is Jev cheaper than GPT-class models for ticket routing?
- In one measured test on 791 labeled decisions, Jev cost about 1.5 cents per 1,000 8-way routing decisions, against about 7 cents for GPT-5.4 nano and 8.7 cents for Gemini 3.5 Flash-Lite.
- How many queues can Jev choose between?
- A Choice question accepts up to 255 options. Each option adds to the cost, so list only the queues you actually staff.
[ For developers ]Request body, wiring and pitfalls
The request
One request per ticket with all 3 questions batched over the same state, so Jev reads the state once. Keys below are the names you'll see in the response.
{
"model": "jev-latest",
"state": "Ticket #88213 — Plan: Business\n\"I've been locked out since the SSO change this morning. Three people on my team can't log in and we have a client demo at 2pm.\"",
"questions": {
"queue": {
"type": "choice",
"instructions": "Which team should handle this ticket?",
"criteria": {
"billing": "Payments, invoices, refunds, plan changes",
"account_access": "Login, SSO, passwords, permissions",
"bug": "Something in the product is broken",
"how_to": "Questions about using a feature",
"feedback": "Feature requests and general feedback"
}
},
"frustration": {
"type": "score",
"instructions": "How frustrated is the customer?",
"criteria": [
"Calm",
"Mildly annoyed",
"Frustrated",
"Angry or threatening to leave"
]
},
"blocking": {
"type": "noul",
"instructions": "Is the customer completely blocked from using the product?"
}
}
}How to wire it up
- 1.Call Jev when the ticket is created, before any agent sees it.
- 2.Assign the queue automatically when confidence ≥ 0.7; leave the rest in a triage queue.
- 3.Raise priority when blocking ≥ 0.6 or frustration ≥ 2.5.
- 4.Log Jev's answer next to the queue the ticket was finally closed in — that's your free accuracy dataset.
Watch out for
- Overlapping queues ("technical" vs "bug") confuse people and Jev alike. Make criteria mutually exclusive.
- Choice questions with many options add tokens per option. 77-way routing cost ~2.6× more than 8-way in the same test.
New to writing Jev questions? Read the question design guide.
Related use cases
Can Jev moderate user content and catch prompt injection?
Flag spam, abuse and prompt injection before content goes live.
Can Jev filter papers, articles and news for relevance?
Surface only the papers and headlines that match what you care about.
Can Jev classify keywords by search intent?
Label keyword lists by intent and relevance, at any volume.