User Intent Classifier: understand the next step
Identify Buy, Get help, Cancel, or Learn more intent in customer messages. Review cancellation language in the context of the conversation.
Free preview limits apply. Sign in to use your credits and keep a decision history.
Your next decision starts here.
Add context above, or try an example to get started.
A mentioned action is not always a requested action
A customer asking about cancellation terms may still be evaluating a purchase. Someone comparing prices may need information rather than a sales handoff. Intent classification is useful when it preserves that distinction. Use the latest message with enough conversation context to understand what is being requested now. The result is a suggestion for the next response; it is not consent to cancel a subscription, send an offer, or complete a purchase on the customer’s behalf.
Work through the cases
Illustrative inputs and review guidance, not recorded model outputs. Run your own context to inspect a decision.
Learning about cancellation
Before I subscribe, can you explain how cancellation works?
The customer is asking for policy information before buying. Mentioning cancellation does not establish a present request to end a subscription. Under an explicit-intent policy, examine Learn more and preserve the pre-purchase context instead of routing solely on the word “cancellation”.
A comparison without a decision
I found a cheaper service. Can you explain what I would lose if I switched?
This is a comparison request and may signal uncertainty, but it is not necessarily a cancellation instruction. Answer the concrete question. A downstream retention workflow should require its own eligibility and consent rules rather than automatically issuing a discount based on a label.
The last message changes meaning with history
Earlier: “Please end my subscription at renewal.” Latest: “Yes, that is what I want.”
The last sentence alone is underspecified. Relevant history clarifies what “that” refers to. Retain who said what and the latest correction; a long transcript with an old cancellation request can be misleading if the customer has since changed their mind.
Set the boundaries before you classify
Suggested review rules for these options. Include the rules relevant to your team in the question or context when you run a decision.
| Choice | Use it when | Check the boundary |
|---|---|---|
| Buy | The customer expresses a wish to purchase, subscribe, or expand. | A general price comparison can still be information gathering. |
| Get help | The customer seeks resolution of an existing product or account issue. | A failure complaint is not automatically a request to leave. |
| Cancel | The customer expresses a current intention to end the service. | Policy questions, quoted text, and hypothetical cancellation are different. |
| Learn more | The customer requests information without a clear transaction or support action. | Look for a concrete commitment elsewhere in the relevant conversation. |
Where to take a closer look
Negation and a change of mind
“I no longer want to cancel” and “I do not think I want to continue” require different responses. Include such cases in a reviewed dataset and preserve the newest statement rather than treating every cancellation keyword as equivalent.
Old intent overriding the current request
When summarizing a conversation, keep chronology and speaker attribution. An agent’s suggested option is not the customer’s choice. Reclassify when new information arrives instead of retaining one label for the entire relationship.
Separate response routing from irreversible actions
Review false Cancel labels and missed cancellation requests separately. Their costs depend on what follows the label: preparing a helpful response is different from sending an unsolicited offer or changing billing. Evaluate the exact next step you plan to automate.
Use reviewed conversations to select thresholds for that workflow. Keep account changes behind explicit confirmation and established cancellation controls, regardless of confidence. If intent remains ambiguous, ask a focused question rather than assuming either purchase or cancellation.
A Jev API example for your next integration
Call the iajev REST API with cURL. Receive a choice, probabilities, and confidence as JSON.
curl https://iajev.com/api/v1/decisions/run \
-H 'Authorization: Bearer YOUR_API_KEY' \
-H 'Idempotency-Key: YOUR_UNIQUE_REQUEST_ID' \
-H 'Content-Type: application/json' \
--data '{
"state": {
"message": "How do I cancel my plan before it renews on the 3rd?",
"plan": "starter"
},
"decision": {
"type": "choice",
"question": "What does this person want to do?",
"choices": [
"Buy",
"Get help",
"Cancel",
"Learn more"
]
}
}'Questions about this tool
Should I send the last message or the whole conversation?
Send the smallest relevant context that resolves references and intent changes. Include the latest message, the action being discussed, and any recent correction. Unrelated history adds noise and may expose details that are unnecessary for the decision.
Does intent classification replace keyword rules?
It can complement them. Keep deterministic controls for explicit cancellation buttons and account actions. Use classification for ambiguous natural-language messages, and compare its mistakes with your existing rules before changing a live routing workflow.
Using iajev
Add context
Include the facts needed for this decision.
Review the choices
Customize the question and options to match your workflow.
Run and review
Inspect the distribution before taking action.
Is confidence the probability that a decision is correct?
No. Confidence is a separate model estimate; the choice probabilities describe the distribution over your options. Neither is a measured accuracy guarantee.
Can I compare two inputs?
Edit the context and run the same question again to compare probabilities. Changing the question or choices starts a new comparison.