Preflight
AI triage for airline crew requests
Powered by Jev · TypeSafe AI
Paste a message from an airline crew member. In well under a second it decides how urgent it is and which department should handle it, with the probabilities behind the decision.
Or try an example
Where this could be used
Ideas for a classifier like this in front of a crew support desk. This demo handles one message at a time and isn’t connected to any mailbox.
Inbox · concept
A mailbox of four crew requests. Before triage they sit in arrival order. After triage the same messages are tagged with urgency, department and tone, and the most urgent one moves to the top.
Smarter inbox
Tag incoming crew emails in Outlook with urgency, department and tone, so handlers know where to start and what a message is about before they open it.
Escalation · concept
Three example signals and where each would be sent: a safety concern goes to Flight Safety, a medical concern to Crew Medical, and misconduct or harassment to HR and Contracts.
Escalation
Send safety, medical and misconduct signals straight to the right desk instead of leaving them in a shared inbox.
Queue · concept
Four requests arrive in time order, then sort themselves with the most urgent first. A message that is too unclear to classify confidently is set apart in a human review lane.
Urgent first
Put same-day and critical requests at the top of the queue, with a review lane for unclear or low-confidence messages.
Insights · sample data
Sample data, not real results. Requests this week by department: Crew Control 42, Crew Planning 31, Payroll 27, Crew Medical 12. Frustrated messages rose from 9 to 15 percent. Missing payslip messages per week: 3, 4, 3, 5, 4, 6, 12, 17, a recent spike.
Spot patterns
Track what crew are asking about and how they feel, to catch recurring problems such as missing payslips or hotels early.
How it works
Jev is a “System One” model: it doesn’t write text, it answers typed questions. Each request asks it a choice for the department, a score for the urgency, a choice for the tone, and several yes/no signals such as medical or safety concerns, all in a single call. Answers come back as calibrated probabilities, so low-confidence cases can be sent to a human instead of guessed.
Everything here is invented sample data. Please don’t enter personal or real crew information.
Contact
Questions, ideas or feedback? I’d love to hear them.