# Recipe: route and prioritise support tickets > Pick the team, rate the urgency and check for churn risk in one call, then route the ticket or hand it to a person when an answer abstains. Every incoming ticket gets a team, a reply deadline and a churn flag before anyone opens it. One request answers all three questions, and the ticket text is billed once. ## The request - **State.** The ticket as JSON, so each question can point at `{{ticket.body}}` while the model can still read the channel, the subject and the customer fields. Send only the fields the decision needs: no email addresses, no account numbers. - **`team` is a `pick`.** The teams are mutually exclusive, so one pick is better than four checks. The option descriptions say what each team does; `other` needs no description. - **`urgency` is a `rate`** on four levels from lowest to highest. `rate` is in beta, so its floor is low and the code below only uses it to raise a priority. - **`churn_risk` is a `check`**: one yes or no fact, answered as the probability of yes. - **`min_confidence`** on every question your code acts on. Below it the answer abstains, and the ticket goes to a person. ```json { "state": { "ticket": { "channel": "chat", "subject": "App logs me out every few minutes", "body": "Since this morning your app logs me out every few minutes and I lose the order I am building. I run a small bakery and must order flour before 14:00 today. If this keeps happening I will move to another supplier." }, "customer": { "plan": "business", "customer_since": "2024-03-02", "open_tickets": 1 } }, "questions": { "team": { "type": "pick", "instructions": "Which team should handle {{ticket.body}}?", "options": { "billing": "Payments, refunds and invoices", "technical": "Bugs, outages and sign-in problems", "shipping": "Delivery and returns", "other": null }, "min_confidence": 0.5 }, "urgency": { "type": "rate", "instructions": "How fast must we reply to {{ticket.body}}?", "levels": ["Can wait", "This week", "Within two working days", "Today"], "min_confidence": 0.3 }, "churn_risk": { "type": "check", "instructions": "Does the writer of {{ticket.body}} say they may cancel or move to another supplier?", "min_confidence": 0.6 } } } ``` ## Run it Put a test key in `DEX_API_KEY` (see the [Quickstart](/docs/quickstart/#get-a-key-in-60-seconds)). Save the Python code as `support_routing.py` and run `python support_routing.py`. Save the TypeScript code as `support-routing.mts` and run `npx tsx support-routing.mts`: the code uses `await` at the top level, and the `.mts` ending makes the file an ES module. Install the SDKs from [Downloads](/docs/reference/sdks/#downloads). ```bash tab="curl" curl https://api.thinqit.ai/v1/decide \ -H "authorization: Bearer $DEX_API_KEY" \ -H "content-type: application/json" \ --data-binary @- <<'DEX_REQUEST' { "state": { "ticket": { "channel": "chat", "subject": "App logs me out every few minutes", "body": "Since this morning your app logs me out every few minutes and I lose the order I am building. I run a small bakery and must order flour before 14:00 today. If this keeps happening I will move to another supplier." }, "customer": { "plan": "business", "customer_since": "2024-03-02", "open_tickets": 1 } }, "questions": { "team": { "type": "pick", "instructions": "Which team should handle {{ticket.body}}?", "options": { "billing": "Payments, refunds and invoices", "technical": "Bugs, outages and sign-in problems", "shipping": "Delivery and returns", "other": null }, "min_confidence": 0.5 }, "urgency": { "type": "rate", "instructions": "How fast must we reply to {{ticket.body}}?", "levels": ["Can wait", "This week", "Within two working days", "Today"], "min_confidence": 0.3 }, "churn_risk": { "type": "check", "instructions": "Does the writer of {{ticket.body}} say they may cancel or move to another supplier?", "min_confidence": 0.6 } } } DEX_REQUEST ``` ```python tab="Python" # Save as dex_request.py, then run: python dex_request.py import json from thinqit_dex import Client client = Client() # reads DEX_API_KEY from the environment request = json.loads(r''' { "state": { "ticket": { "channel": "chat", "subject": "App logs me out every few minutes", "body": "Since this morning your app logs me out every few minutes and I lose the order I am building. I run a small bakery and must order flour before 14:00 today. If this keeps happening I will move to another supplier." }, "customer": { "plan": "business", "customer_since": "2024-03-02", "open_tickets": 1 } }, "questions": { "team": { "type": "pick", "instructions": "Which team should handle {{ticket.body}}?", "options": { "billing": "Payments, refunds and invoices", "technical": "Bugs, outages and sign-in problems", "shipping": "Delivery and returns", "other": null }, "min_confidence": 0.5 }, "urgency": { "type": "rate", "instructions": "How fast must we reply to {{ticket.body}}?", "levels": ["Can wait", "This week", "Within two working days", "Today"], "min_confidence": 0.3 }, "churn_risk": { "type": "check", "instructions": "Does the writer of {{ticket.body}} say they may cancel or move to another supplier?", "min_confidence": 0.6 } } } ''') decision = client.decide( request["state"], request["questions"], ) for question_id, answer in decision.answers.items(): print(question_id, answer) ``` ```ts tab="TypeScript" // Save as dex-request.mts, then run: npx tsx dex-request.mts (Node.js 18 or newer) import { Client, parseRequest } from "@thinqit/dex"; const client = new Client(); // reads DEX_API_KEY from the environment // parseRequest keeps the key order of the text (JSON.parse would move labels such as "1" to the front). const request = parseRequest(`{ "state": { "ticket": { "channel": "chat", "subject": "App logs me out every few minutes", "body": "Since this morning your app logs me out every few minutes and I lose the order I am building. I run a small bakery and must order flour before 14:00 today. If this keeps happening I will move to another supplier." }, "customer": { "plan": "business", "customer_since": "2024-03-02", "open_tickets": 1 } }, "questions": { "team": { "type": "pick", "instructions": "Which team should handle {{ticket.body}}?", "options": { "billing": "Payments, refunds and invoices", "technical": "Bugs, outages and sign-in problems", "shipping": "Delivery and returns", "other": null }, "min_confidence": 0.5 }, "urgency": { "type": "rate", "instructions": "How fast must we reply to {{ticket.body}}?", "levels": ["Can wait", "This week", "Within two working days", "Today"], "min_confidence": 0.3 }, "churn_risk": { "type": "check", "instructions": "Does the writer of {{ticket.body}} say they may cancel or move to another supplier?", "min_confidence": 0.6 } } }`); const decision = await client.decide(request); console.log(decision.answers); ``` ## Expected output ```json { "id": "req_01M3JE2D3792VXG4CSFWBBBC68", "object": "decision", "created": 1790546359, "model": "dex-1.0.1", "served_by": "gpu", "calibration": "cal-20260926-1", "answers": { "team": { "type": "pick", "choice": "technical", "probabilities": { "billing": 0.0392, "technical": 0.9066, "shipping": 0.0299, "other": 0.0243 }, "confidence": 0.8673, "abstained": false }, "urgency": { "type": "rate", "rating": 2.6751, "levels": ["Can wait", "This week", "Within two working days", "Today"], "probabilities": [0.0335, 0.058, 0.1083, 0.8002], "confidence": 0.5117, "abstained": false }, "churn_risk": { "type": "check", "probability": 0.9604, "confidence": 0.9209, "abstained": false } }, "usage": { "input_tokens": 185, "state_tokens": 99, "question_tokens": 86, "allowance_tokens": 0, "paid_tokens": 0, "charge_micro_cents": 0, "unit_price_micro_cents": 0, "tier": "test" } } ``` Captured from the live API on 2026-09-27 with a test key: model `dex-1.0.1`, calibration `cal-20260926-1`, `served_by: gpu`, 185 input tokens (99 for the state, 86 for the questions). A test key is charged nothing, so `tier` is `test` and the charge is 0. On this exact version the same request always returns these answers. | Question | Type | Answer | Confidence | `min_confidence` | Abstained | | --- | --- | --- | --- | --- | --- | | `team` | pick | `technical` (0.9066) | 0.8673 | 0.5 | no | | `urgency` | rate | rating 2.6751, most likely `Today` (0.8002) | 0.5117 | 0.3 | no | | `churn_risk` | check | yes with probability 0.9604 | 0.9209 | 0.6 | no | The ticket goes to `technical` with probability 0.9066. The urgency rating of 2.6751 sits close to `Today`, and the writer says they may move to another supplier (0.9604). No answer abstained, so the code below can act on all three. ## Act on it The functions `route_to_person`, `route_to_team`, `set_priority` and `flag_for_retention` stand for your own code. ```python tab="Python" team = decision.pick("team") if team.abstained: route_to_person(ticket_id) # two teams are close: a person decides else: route_to_team(ticket_id, team.choice) urgency = decision.rate("urgency") if not urgency.abstained and urgency.rating >= 2.5: set_priority(ticket_id, "high") churn = decision.check("churn_risk") if not churn.abstained and churn.probability >= 0.5: flag_for_retention(ticket_id) ``` ```ts tab="TypeScript" const { team, urgency, churn_risk: churn } = decision.answers; if (team?.type === "pick") { if (team.abstained) routeToPerson(ticketId); // two teams are close: a person decides else routeToTeam(ticketId, team.choice); } if (urgency?.type === "rate" && !urgency.abstained && urgency.rating >= 2.5) setPriority(ticketId, "high"); if (churn?.type === "check" && !churn.abstained && churn.probability >= 0.5) flagForRetention(ticketId); ``` `parseRequest` keeps the request as text, so the TypeScript answers have the general `Answer` type and are narrowed on `type`. Build the questions with the SDK's `pick`, `rate` and `check` helpers instead (as in the [Quickstart](/docs/quickstart/#make-your-first-call)) and `decision.answers.team.choice` is typed as your labels. ## Adapt it - **Your own teams.** Up to 255 options per pick. Describe each team by what it handles, not by its internal name. - **A deadline, not a feeling.** When you have a service level, compute the deadline in your code and pass it as a state field; Dex reads, it does not calculate. See [Known limits](/docs/concepts/known-limits/#date-and-amount-arithmetic-is-unreliable). - **Text from customers can try to steer the classifier.** For a high-impact action, add a guardrail check as in the [agent tool gate recipe](/docs/cookbook/agent-guardrails/). - **Tune the floors on your own tickets.** Run a few hundred past tickets with a test key, compare with the team that handled them, and raise `min_confidence` until the automatic routes are right often enough. See [Abstention](/docs/concepts/abstention/).