Incoming queue // taxonomy under pressure
INTENT
TETRIS
Customer messages are falling. Route each one to the right intent before it hits the queue. Sounds easy until the boundaries start to blur.
Free to play. Nothing is saved. No account required.
SCORE0000
STREAK×0
ROUND01
QUEUE HEALTH♥ ♥ ♥
OPERATIONS TRAINING MODULE 01
KEEP THE QUEUE CLEAN.
Read the message. Choose its primary intent. Use keys 1 through 4 or select a route below.
The correct answer is the customer's primary goal, not every topic their message happens to mention.
The serious idea inside the game
Intent classification is a boundary problem.
What is an intent?
In conversational AI, an intent represents the goal behind a user's message. “Where is my order?” maps to delivery tracking because the customer wants an arrival update, even though the message also refers to an order.
Why do boundaries matter?
A useful taxonomy makes intents mutually exclusive enough to route reliably. If billing dispute and refund request accept the same language, the classifier has to guess and customers land in the wrong workflow.
What makes a message hard?
Real customers combine context, emotion, and multiple needs. A strong system identifies the job to be done, handles secondary needs, and knows when confidence is too low for automatic routing.
TRY THIS ONE
“I returned the damaged shoes last week. Why am I still being charged?”
PRIMARY: BILLING
The return is context. The unresolved charge is the goal. Routing only on the word “returned” would send this customer to the wrong workflow.