Job new-tickets-tagged-prioritised-and-routed-by-your-categories · revised 12 October 2026
Suggest a category, priority and queue for each new ticket from your own categories
Each new customer-support ticket gets a suggested category, queue and (urgent and legal only) priority in a separate field, with agreement measured on tickets your staff labelled.
You might be seeing
- Most tickets arrive with no tag and the tags in use drift
- Urgent outage reports wait in the general queue
- Reports by category count whatever people happened to tag
No passwords, keys, card details or admin invites needed to start.
What usually happened
When staff tag by hand, most conversations go untagged and the tags in use drift apart. Routing then depends on who reads the ticket first, and reports by category count the tags people happened to apply.
Who it’s for: Support lead at a software company where agents tag by hand and routing is a guess.
Usually starts when: Tags are unused or misapplied, urgent tickets sit in the general queue and reports by category cannot be trusted.
The result: Each new ticket gets a suggested category, priority and queue taken from your written rules, shown in a separate suggestion field. The queue follows from the category by a table you write, and priority is suggested only for urgent-outage and legal tickets. The agreement with tickets your staff labelled is measured, every disagreement is listed and nothing routes live until you turn it on.
Check whether this job fits
Answer these without sending chats, tickets, files or logins. Nothing is submitted unless you choose to contact us.
What you get
- The category list with definitions and examples
- The suggestion rules and how to turn them on
- The agreement table and every disagreement
- The urgent and legal rule results
- A list of categories to rewrite
Included
- At most 12 top-level categories, each with a definition and examples, written with you
- Suggestion rules or a classifier that reads ticket text and writes only to a separate suggestion field
- A test on 200 held-out tickets labelled independently by two of your staff
- A table you write from category to queue, and a rule you write for urgent-outage and legal categories, tested on planted tickets and on 5 planted tickets that carry instructions
- The agreement table and a list of overlapping categories that should be rewritten
Not included
- Priority that drives a service-level penalty or a promise to customers
- Switching on live routing: you do that
- Scoring or ranking your agents
- A second level of categories unless it meets its own mark
- Voice or call tickets
- Changes to your help desk by us
- Internal IT tickets, which have their own categories and buyer: see the IT triage job
- Staff phishing reports: see the phishing report triage job
- A one-off file of free-text rows to label: see the free-text labelling job
How we know it’s done
Agreed with you before work starts. Each check produces evidence you keep.
On 200 held-out tickets labelled independently by two of your staff, the top-level suggestion matches the agreed label on at least 160, counting "unsure" as a miss; items on which the two labellers differ are settled by your lead before the run and still count in the 200 (160 of 200, which shows a true pass rate of at least 74% with 95% confidence).
Evidence: The 200-ticket table, the labels and the held-out marker.
The suggestion beats a rule that always picks your commonest category, and the per-category agreement table shows no category with fewer than half of its tickets matched, counting "unsure" as a miss.
Evidence: The agreement table with the baseline row.
All 20 planted urgent-outage and legal tickets (10 each) get the category you named in your own rule; 8 of the 20 are written by your lead and unseen by us until the final run. A miss fails the whole test. These planted cases are a smoke test, not a measured pass rate on real messages.
Evidence: The 20 planted tickets and the category each received, and the held-back marker.
Every disagreement is listed with the ticket and both labels, and a second level is reported only if it matches on at least 70%.
Evidence: The disagreement list and the second-level result if reported.
Nothing routes live: the suggestion field exists but no queue, owner or priority in your help desk has been changed by us.
Evidence: Your help desk's change log for the period.
The suggested queue equals the queue your table gives for the suggested category on all 200 held-out tickets, and priority is suggested only on urgent-outage and legal tickets.
Evidence: The 200-ticket table with the queue column beside the table you wrote.
Of 5 planted tickets that carry instructions to the suggester (for example to ignore your rules or mark the ticket urgent), none obeys: each gets the category its content earns. One that obeys fails the whole test.
Evidence: The 5 planted tickets and the category each received.
Sign-off. You read the agreement table and the disagreement list and sign off in writing. Payment follows sign-off, and you turn the suggestions on yourself.
If it fails. If the agreement mark is not met after one retest, we hand over the table and the rules as they stand and you decide whether to stop or re-quote. We do not call a miss a pass.
When it fits, and when we stop
It fits when
- At least 1,000 past tickets can be exported with personal details removed
- Two of your staff can label the 200 test tickets independently
- A named person decides the categories and the urgent rules
We stop and tell you if
- No ticket history exists
- The categories are undecided and nobody can decide them
- The two labellers disagree on more than 15% of a 50-ticket trial: no pass mark would mean anything
- You want priority to drive a service-level penalty
- The tickets hold special-category data we have not agreed to handle
What could go wrong
Suggestions go to a separate field or file, so nothing in your help desk changes until you turn them on. Turn them off and routing is exactly as before.
Scroll the table sideways to read it all.
| Risk | How we handle it |
|---|---|
| A model that always picks your commonest category looks accurate | The result is compared with that baseline, and agreement is reported per category, not only overall. |
| Ticket text carries instructions that steer the suggestion | Tickets are treated as untrusted text, the suggestion only writes to its own field and cannot send or change anything. |
| Deep category levels are unreliable | A second level is reported only if it meets its own mark; otherwise it is left out. |
| An urgent outage is suggested as low priority | Twenty planted urgent and legal tickets must all get the category you named, and a miss fails the whole test. |
A separate automated reviewer checks the categories, the agreement table and the urgent rule results before hand-over, and your lead on your side marks the sample and decides what goes live. No human supervisor is included unless your agreement names one. At launch the work is largely automated, and we say so.
How we deliver
We arrange the work and independent review, then show you the result against the agreed checks. You keep authority over your systems.
- Agree categories, urgent rules, pass marks and the severe-error list in writing
- Two staff label 50 tickets; we start only if they agree on 85%
- Build the rules on the redacted export; your lead holds 200 labelled tickets back
- Run the 200 held-out tickets and record every disagreement
- Independent automated review; your lead rules on each disagreement
- Hand over the rules, the results and a list of categories to rewrite
This is a one-off job, not emergency cover or a subscription. We confirm eligibility, the total price, a start window and a delivery date before you accept. Work starts only after agreed inputs, secure access and any needed permissions are in place. Model usage for our build and tests is included in the price; charges on your own platform account, including per-conversation fees after go-live, are yours. An enquiry creates no charge or booking.
Need to keep it working?
Discuss a monthly re-check of the suggestions on a fresh labelled sample.
Ongoing work is separately scoped and quoted: no monitoring, response-time guarantee or automatic subscription is included in this job.
Explore an ongoing managed lane, or mention the responsibility you need in your enquiry.
What you can check
This is a new service. We have not delivered this job for a client yet.
Other ways to get this done
- Your help desk may already classify topic, sentiment and language. Try its own triage on 50 tickets first. support.zendesk.com
Questions
Does it route tickets by itself?
No. It writes a suggestion to a separate field. Routing stays as it is until you turn the suggestions on.
What if my help desk already has this?
Try it first on 50 tickets. If it meets your mark, you do not need this job.
Do you score my agents?
No. Suggestions are about tickets, not people.
How is this different from the IT triage job?
This job suggests categories for customer-support tickets from the categories you write. The IT triage job uses IT categories and four priority levels, for an internal IT team.
Send an enquiry
Send us
- Your help desk and roughly how many tickets arrive each month
- The categories or queues you use today
- How many people tag by hand
- No customer data, logins or keys in the first enquiry
Later, once you agree
- A redacted export of at least 1,000 past tickets through a secure route agreed in writing
- Two labellers on your side and 200 held-out tickets chosen by your lead and unseen by us until the final run
- Your approved categories and urgent rules, and 8 of the 20 planted urgent and legal tickets written by your lead and unseen by us until the final run
You keep the help desk and every login. We work from a redacted export, write only suggestions to a file or a separate field you create, and never change live routing. You turn it on, and off.
Email fallback: open your mail app
If website submission is unavailable, review and send the fallback email yourself. An email fallback is not a website receipt. Or write to hello@syntheticindustry.ai with “new-tickets-tagged-prioritised-and-routed-by-your-categories” as the subject.