What detection looks for
Commitments usually combine a speaker, a future action and an expectation: “I’ll send,” “we will confirm,” or “I’ll check and come back to you.” A time phrase strengthens the signal but is not always present.
What should not count
A suggestion, historical statement or customer request is not necessarily an agent commitment. “You can try again tomorrow” differs from “I’ll check this again tomorrow.” Review controls are essential.
Promise output
A Promise detection can include whether a promise exists, summary, original text, due phrase, due date, type and confidence. A person can then confirm or correct it.
How Promise handles this in Zendesk
Promise is built around the commitment as a separate operational object. It keeps the original wording, a concise summary, owner, due date, status and source ticket together. Agents can review and correct the record. Managers can see team status and open the underlying ticket before acting.
Promise costs $5 per agent/month and is not yet live on the Zendesk Marketplace. AI results can be incomplete or incorrect, so the product supports human review rather than claiming perfect detection.
Examples from real support work
I’ll check with the warehouse and update you tomorrow.
I’ll ask engineering and come back to you Friday.
I’ll confirm with the hotel and update you this afternoon.
I’ll contact the carrier and send the latest status tomorrow.
Questions about detect agent promises in Zendesk
Can AI detect multiple promises in one reply?
Promise is designed to support multiple commitments in a reply.
Does every future-tense sentence count?
No. The sentence must represent an agent or team obligation to the customer.
Is detection fully automatic?
Detection can be automated, but important results should be reviewed by a person.