Use case
Feedback and NPS calls that capture the reason behind the score
A short call after a visit, delivery, class or service. The AI agent asks for a score, asks why, and saves both in the caller's own words. Unhappy customers are assigned to your team as soon as the call ends.
- AI agentHello Priya ji, this is Riya, an AI assistant from Greenleaf Home Store, calling about the sofa delivered on Tuesday. Do you have two minutes for some quick feedback?
- CallerSure, go ahead.
- AI agentOn a scale of 0 to 10, how likely are you to recommend us to a friend?
- CallerMaybe a 6.
- AI agentWhat is the main reason for that score?
- CallerThe sofa is nice, but the delivery came two days late and nobody called to tell me.
- AI agentI'm sorry about that. Would you like someone from our delivery team to call you about it?
- CallerYes, that would help. After 6 pm is best.
- AI agentThank you, Priya ji. Is there anything else you would like us to do better?
- CallerNo, that's it.
Survey links get ignored and scores come without reasons
A survey link on WhatsApp or email gets a few replies, often from people who are either delighted or upset. Many of the people in between never open it.
When a manager calls customers personally, the answers are richer, but only a handful of calls get made each week and the notes are hard to compare.
A short, consistent call asks every customer the same questions in the same order and saves the answers side by side. An unhappy customer on the phone can also be passed to a person while it still matters.
On the call
What the agent does, step by step
- 1
Call after the service
A workflow calls the customer after the visit, delivery or class, triggered by your system through a webhook. Or run a campaign over the week's list.
- 2
Short, specific opening
The agent says who is calling, what the call is about, and that it will take two minutes.
- 3
Score first
It asks for a 0 to 10 score, or whichever scale you use.
- 4
Then the reason
It asks what drove the score, with one follow-up at most. Script it not to argue; it does not read the answer back.
- 5
Route the unhappy ones
Low scores and complaints get your disposition and go to a person, with the summary and the reason.
Into your CRM
What it captures on every call
- Score (0 to 10, or your own scale)
- Main reason for the score, in the customer's words
- Specific complaint, if any (delivery, staff, product, billing)
- Whether they want a call back from your team
- Suggestion for improvement
- Whether they would buy or visit again
After the call
What happens next
| When | Then |
|---|---|
| High score, no issues | Marked with your Promoter disposition and stopped. If you offer a review link and they say yes, that WhatsApp template is sent. |
| Low score or a complaint | Assigned to a person on your team with the summary and the reason, so they call back knowing the problem. |
| The customer asks for a call back | Saved as a callback request on the call record, with what they said ('after 6 pm') in the transcript. Set Callback to assign so a person calls back. |
| No answer | Workflow calls are retried after your gap, up to your maximum attempts, then stopped; a weekly campaign doesn't re-dial, so re-run it over those customers. Calls where nobody spoke are marked Incomplete and carry no score. |
Set it up
- Pick the scale (0 to 10 NPS, 1 to 5, or yes/no), put the score and reason questions in the script, and add them as extraction questions so the analysis saves the answers.
- Keep the script to three or four questions and about two minutes.
- Add dispositions such as Promoter, Passive, Detractor and Complaint. Set Promoter and Passive to stop, and assign Detractor and Complaint to a person.
- Trigger calls from a workflow after the service, or run a weekly campaign over a list.
- Choose a calm voice and pace, and check it with the voice preview.
- Optionally link an approved WhatsApp template with your review link for happy customers who agree to it.
What to measure
- Response rate: completed surveys ÷ customers called
- Score distribution and NPS, by branch, product or team
- Top reasons behind low scores, from the captured answers
- Time from a low score to a call back from your team
Where teams use it most
Questions about feedback surveys
Is a survey call better than a WhatsApp survey link?+
They do different jobs. A link is cheap and works for customers who reply. A call gets a score and a reason from people who would never open the link, and lets an unhappy customer ask for help there and then. You can use both, and call the people who ignored the link.
Will the agent argue with a customer who gives a low score?+
It shouldn't, and you control that in the script. We recommend a script that thanks the customer, asks for the reason, offers a call back from your team and moves on. The agent does not read the complaint back to them.
How are open answers stored?+
Extracted answers keep only what the caller actually said, next to the score, a short summary, the recording and the full transcript. You can search, filter and export the call log.
How long should a feedback call be?+
Two to three minutes is enough for a score, a reason and one follow-up. Each agent also has a maximum call length, six minutes by default, which you can change.
Can some customers get the survey in Hindi and others in English?+
Yes. Hindi, English and Hinglish are all in production. Make one agent per language and run each over its own list.
Hear it on a real phone call.
Book a 20-minute demo and we'll build a first agent around your script, or ring our test line and talk to one right now.