Find out how each support conversation ended and what is still owed
Reads a finished conversation and records whether it was resolved, how the customer felt, complaints, threats to leave, open promises and the next action, so managers see outcomes rather than ticket counts.
Try it on this example
Channel (email, web chat, chat app, SMS, call): Web chat
The whole conversation, oldest message first
- At the end of the conversation, what is the state of the customer's request?In progress76%
- How satisfied does the customer say they are with how the contact was handled?Very dissatisfied94%
- Does the customer express dissatisfaction with the service, the product or how they were treated?Yes98%
- Does the customer say they contacted the company before about this same unresolved issue?Yes98%
- Does the customer say they intend to close their account, leave or stop using the service?No87%
- Does the customer ask for a manager, a supervisor or the formal complaints process?Yes99%
- Did an agent make a specific promise that the conversation does not show as kept?Yes85%
- How did the customer's tone change from the start of the conversation to the end?Worsened99%
- Which single action should the company take next?Escalate87%
These are real answers stored from one run on this example.
The prism behind it
Find out how each support conversation ended and what is still owed
Fields
- Channel (email, web chat, chat app, SMS, call)
- The whole conversation, oldest message first
Context
The text is one support conversation between a customer and the company, oldest message first. It may carry sender names and times, or no speaker labels at all, as in a call transcription; the customer is the person who contacted support. Judge only what the conversation shows, as it stands at its last message. Counts, rates and response times are worked out elsewhere from these answers and the timestamps.
Questions
At the end of the conversation, what is the state of the customer's request? Choice
How satisfied does the customer say they are with how the contact was handled? Scale
Judge the customer's own words about the handling, not their feelings about the underlying problem, and not your view of the agent.
Does the customer express dissatisfaction with the service, the product or how they were treated? Yes / No
Reporting a fault neutrally is not in itself a complaint. Yes: The customer expresses dissatisfaction. No: The customer reports or asks without expressing dissatisfaction.
Does the customer say they contacted the company before about this same unresolved issue? Yes / No
Yes: The customer refers to an earlier call, chat, email or ticket about this same issue that did not resolve it. No: No such earlier contact is mentioned, or it was about something else.
Does the customer say they intend to close their account, leave or stop using the service? Yes / No
Exclude intentions that are hypothetical, conditional on something that has not happened, or withdrawn later in the conversation. Yes: The customer states an actual intention to leave or close the account. No: No such intention appears, or it is hypothetical, conditional or withdrawn.
Does the customer ask for a manager, a supervisor or the formal complaints process? Yes / No
Yes: The customer explicitly asks to escalate or for someone more senior. No: No such request appears.
Did an agent make a specific promise that the conversation does not show as kept? Yes / No
Promises include calling back, sending something, checking or investigating, passing to a team, or paying or refunding. Yes: A specific agent promise appears and nothing in the conversation shows it was kept. No: No specific promise was made, or every promise is shown as kept.
How did the customer's tone change from the start of the conversation to the end? Choice
Which single action should the company take next? Choice
Look at where the conversation has got to by its end and choose the action that should happen first.
Lens columns
resolution, resolution_probability, customer_satisfaction, customer_satisfaction_average, complaint, complaint_probability, repeat_contact, repeat_contact_probability, leaving_intent, leaving_intent_probability, escalation_request, escalation_request_probability, open_commitment, open_commitment_probability, sentiment_change, sentiment_change_probability, next_action, next_action_probability
Run it on your own text
Add this prism in the app, change any question, and test it on a file of your own.