Outbound AI calls for feedback collection
Automatically call customers to collect feedback, NPS scores, and detailed responses. Natural conversation beats robotic IVR surveys with higher completion rates.
Hi, this is Alex calling from CloudTech Solutions. I'm following up on your recent support interaction. Do you have about 2 minutes for a quick feedback call?
Sure, I can do that.
Great, thank you! On a scale of 1 to 5, how would you rate your recent support experience?
I'd say a 3. It was okay.
Thanks for that feedback. A 3 tells me there's room for improvement. What went well during your interaction?
The technician knew what he was doing and fixed my issue eventually.
Good to hear the issue was resolved. What could we have done better?
I was on hold for 45 minutes before getting through. That was frustrating.
I completely understand—45 minutes is too long. I'll make sure this feedback gets to our team. Last question: on a scale of 0 to 10, how likely are you to recommend CloudTech to a colleague?
Probably a 7. The product is good but the support wait times need work.
Thank you so much for your time and honest feedback. We're actively working on reducing wait times. Have a great day!
Configure your AI agent's personality and behavior.
You are conducting a brief customer satisfaction survey for CloudTech Solutions. The customer recently had a support interaction.
Survey flow:
1. Introduce yourself and confirm they have 2-3 minutes
2. Ask about their recent support experience (1-5 rating)
3. Ask what went well
4. Ask what could be improved
5. Ask if they'd recommend us (NPS: 0-10)
6. Thank them and offer to connect them with support if needed
Be conversational, not robotic. If they give a low score, empathize and dig deeper. Keep it brief and respectful of their time.Define the actions your AI agent can perform.
Record a numerical rating from the customer
{
"question_id": string - Identifier for the question (support_rating, nps_score),
"score": number - The rating given (1-5 for satisfaction, 0-10 for NPS),
"verbatim": string (optional) - Any additional comments
}record_rating({"question_id": "support_rating", "score": 4, "verbatim": "The agent was helpful but took a while to resolve"})Record open-ended feedback
{
"question_id": string - Identifier for the question (positive_feedback, improvement_feedback),
"response": string - The customer's verbatim response,
"sentiment": string - positive, neutral, or negative
}record_feedback({"question_id": "improvement_feedback", "response": "Wish the wait time was shorter", "sentiment": "neutral"})Flag this customer for manager follow-up due to negative experience
{
"reason": string - Why follow-up is needed,
"urgency": string - low, medium, or high,
"preferred_contact_method": string (optional) - phone or email
}flag_for_followup({"reason": "Customer very dissatisfied with resolution time, considering cancellation", "urgency": "high"})Mark the survey as complete and calculate scores
{
"completed": boolean - Whether customer completed the full survey,
"partial_reason": string (optional) - If not completed, why
}complete_survey({"completed": true})Handle tool calls from SIP4AI in your backend.
app.post('/webhook/survey', async (req, res) => {
const { tool_name, parameters, call_id, customer_id } = req.body;
switch (tool_name) {
case 'record_rating':
await db.survey_responses.create({
customer_id,
call_id,
question_id: parameters.question_id,
score: parameters.score,
verbatim: parameters.verbatim,
timestamp: new Date()
});
// Real-time alert for low scores
if (parameters.score <= 2) {
await slack.alert(`⚠️ Low score (${parameters.score}) from customer ${customer_id}`);
}
return res.json({ success: true, recorded: true });
case 'record_feedback':
await db.survey_responses.create({
customer_id,
call_id,
question_id: parameters.question_id,
text_response: parameters.response,
sentiment: parameters.sentiment,
timestamp: new Date()
});
// Categorize with AI
const category = await openai.categorize(parameters.response);
await db.survey_responses.update({ category });
return res.json({ success: true, category });
case 'flag_for_followup':
await db.followups.create({
customer_id,
call_id,
reason: parameters.reason,
urgency: parameters.urgency,
status: 'pending'
});
// Notify customer success team
await email.send({
to: 'cs-team@company.com',
subject: `[${parameters.urgency.toUpperCase()}] Follow-up needed`,
body: `Customer ${customer_id}: ${parameters.reason}`
});
return res.json({ success: true, followup_created: true });
case 'complete_survey':
// Calculate NPS and satisfaction scores
const responses = await db.survey_responses.findByCall(call_id);
const nps = responses.find(r => r.question_id === 'nps_score')?.score;
await db.surveys.update(call_id, {
completed: parameters.completed,
partial_reason: parameters.partial_reason,
nps_score: nps,
completed_at: new Date()
});
return res.json({ success: true, nps_score: nps });
}
});Get started with SIP4AI in minutes. Self-host it, connect your SIP line, and deploy your AI agent.