Your customers sign up, start an identity check or setup, then go quiet. Emails go unread, and a small team can't phone everyone. An AI voice agent can call within minutes of the drop-off.

All scripts are illustrative. Names, brands and amounts are placeholders, and each outcome is a result to measure, not a guarantee. Where a script says "I'll text you", your own messaging tool sends that text. Insight7 voice campaigns do not send texts, so connect an SMS tool or have your team send the link.

Four onboarding calls that stop drop-off

Call as soon as possible after the drop-off event. A call triggered by a CRM event reaches the customer while the task is still fresh. A weekly batch call reaches someone who has moved on. Track small wins such as "KYC passed" and "first funding" on their own, so you can tell customer-caused drop-off from product-caused drop-off.

1. Unfinished KYC nudge

Trigger: ID check started, not submitted after 30 minutes.
Agent: "Hi [Name], I'm an AI assistant from [Brand]. Your ID check is nearly done. Can I help with the last step?"
Customer: "The photo keeps failing."
Agent: "Try daylight, no glare. Your team will send the link now."
Handoff: Two failed tries go to a human callback.
Expected outcome: More checks finished. Track KYC completion within 24 hours. In our experience, unfinished KYC later blocks transactions and raises support tickets.

2. Welcome call that confirms sign-up details

Trigger: Account created.
Agent: "Welcome, [Name]. I'm an AI assistant. Is your phone number and address still correct?"
Customer: "The address changed."
Agent: "Thanks. I'll note the new one for review."
Handoff: Name or ID mismatches go to a human.
Expected outcome: Cleaner records and, ideally, fewer failed checks later. Track the share of profiles confirmed in week one.

3. Signed up, not funded

Trigger: Account open, no first deposit or order after 48 hours.
Agent: "Hi [Name], I'm an AI assistant. Anything stopping your first deposit?"
Customer: "I wasn't sure how transfers work."
Agent: "It takes two taps. Want the steps sent to you?"
Handoff: Fee or trust worries go to a person.
Expected outcome: More first deposits. Track signup-to-funded rate, and tag each blocker the customer names.

4. Post-disbursement details check

Trigger: Loan paid out recently, for example the day before.
Agent: "Hi [Name], I'm an AI assistant. Your loan was paid yesterday. Your first repayment is on [date], and the monthly amount is [amount]. Is that clear?"
Customer: "Yes. Can I pay from my wallet?"
Agent: "Yes, the wallet steps will be sent to you."
Handoff: "I didn't take this loan" goes to fraud staff at once.
Expected outcome: Customers who know the terms, and fewer missed first repayments. Track first-payment on-time rate.

Three activation calls that drive first use

Call when a customer has intent but has not yet seen value. Use voice when the message is urgent or may raise questions. Otherwise send a message. If a customer has asked for "no calls", send a message instead.

5. First-feature walkthrough

Trigger: App installed and logged in, no key action after 24 hours.
Agent: "Hi [Name], I'm an AI assistant. You've logged in but not sent money yet. What would you like to do first?"
Customer: "Pay a bill."
Agent: "A link that opens that screen will be sent to you."
Handoff: Any confusion goes to a human callback.
Why here: The customer has installed but not used the product. Do not guide taps during the call. One lender found that walkthroughs kept failing because mobile data cut out when customers used the app mid-call.
Expected outcome: More first actions. Track the share of logins that reach a first action within 48 hours.

6. Pre-due-date repayment reminder

Trigger: Automatic repayment due in two days.
Agent: "Hi [Name], I'm an AI assistant. Your repayment is due on [date]. Is your account funded?"
Customer: "Not yet. Payday is the 4th."
Agent: "Thanks. I'll note it and a reminder will go out on the 4th."
Handoff: Hardship or dispute goes to a human.
Why here: The first successful repayment sets the habit.
Expected outcome: Fewer failed debits. Track the auto-repayment success rate.

7. Stalled setup, find the blocker

Trigger: Setup started, not finished after three days.
Agent: "Hi [Name], I'm an AI assistant. Setup is almost done. What stopped you?"
Customer: "I don't have my document with me."
Agent: "No problem. Want a person to call you tomorrow evening?"
Handoff: Always offer a human callback.
Why here: Customers often stall on small barriers, such as not knowing the app's name, verification issues or older devices.
Expected outcome: Tagged blockers and, ideally, more finished setups. Track completion rate by blocker.

Three retention calls that win customers back

Call lapsed customers while the quiet is still short. Start with dormant reactivation, a common first voice campaign.

8. Dormant customer reactivation

Trigger: No login or transaction for 45 days.
Agent: "Hi [Name], I'm an AI assistant from [Brand]. We haven't seen you in a while. Is there something we could do better?"
Customer: "Fees felt high."
Agent: "Thanks. This month, transfers under [amount] are free. Want the details sent to you?"
Handoff: Complaints about a past issue go to a human.
Expected outcome: Customers who return, plus a tagged reason for the silence. Track reactivated accounts per 100 calls answered.

9. Proactive service check-in

Trigger: An outage ended, a delivery ran late, or an installation finished two days ago.
Agent: "Hi [Name], I'm an AI assistant from [Brand]. Your service was down yesterday morning. Is everything working now?"
Customer: "Yes, but it was slow."
Agent: "Sorry about that. I've logged it for our team."
Handoff: Any unresolved fault goes to a human the same day.
Expected outcome: Problems found before a complaint arrives. Track avoided complaints. Customers want updates about service itself, though some prefer a message to a call.

10. Win-back and churn-reason call

Trigger: Account closed, or no use for 90 days.
Agent: "Hi [Name], I'm an AI assistant from [Brand]. Did you stop using the app? One short reason helps us."
Customer: "Yes. The app kept failing."
Agent: "Thank you. I'll pass that on."
Handoff: Offer a human call to anyone who wants to talk.
Expected outcome: A yes/no plus a short reason. Track tagged churn reasons. Say the brand name first, because lapsed customers often do not recognise the caller and hang up.

How to use these scripts

Pick the script that matches your biggest drop-off, then run it as a campaign. Insight7's outbound voice campaigns are in beta and switched on per organisation. A manager sets a goal, an audience, the AI caller and the rules. Add contacts by CSV upload or type them in.

Set the rules before you launch. The beta spec lists consent marking, calling hours, retry rules and a do-not-call list, so confirm each in the app.

Test first. You can talk to the agent in your browser with no phone number. The dashboard shows calls in progress, and a reviewer can read each transcript when a call ends. Yes/no answers and outcomes are pulled from the transcript. A call where only the AI spoke is not counted as reached, so your reach number stays honest.

Buyers first ask whether this replaces their stack or plugs into it. Ask about your CRM, dialer and SMS tool before you wire the trigger. See the AI voice and chat agents page.

How to adapt them

Swap the trigger and the wait time to match your funnel. Replace placeholders with your brand, amounts and dates. Keep three parts in every script: the AI disclosure in the first line, one question, and a human handoff. Give each script one metric.

Related reading: How to Increase Account Funding Rates for Fintech and Investment Apps With AI Outbound Calls, 7 Signs Your Email and SMS Onboarding Sequence Needs a Voice Channel and How to Notify Customers About Price Changes and New Fees With AI Voice Calls.

Questions people also ask

What are the best AI voice agent use cases for customer onboarding?

Start with calls that follow a clear drop-off event: an unfinished KYC check, a signed-up account with no first deposit, and a stalled setup. Each has a CRM trigger and a measurable result. Add welcome and post-disbursement calls next.

Can an AI voice agent follow up on unfinished KYC?

Yes. The agent can prompt the last step, and your own messaging tool can send a secure link. It should never collect ID numbers or photos by voice. Route repeated failures and mismatches to a human.

Do AI voice calls need customer consent?

In the US, AI-voice calls already need consent under the TCPA, following the FCC's 2024 ruling. Other regions have their own rules. Record consent for every contact. Insight7's beta spec lists consent marking; confirm it in the app.

Should an AI voice agent say it is an AI?

Yes, in the first sentence. EU AI Act Article 50 has applied since 2 Aug 2026 and requires AI that talks to people, including automated phone systems, to say it is AI. The US "say it is AI" rule was still only proposed.

How do you test an AI voice agent before it calls customers?

Talk to it in your browser first, with no phone number needed. Then send a test call to your own phone. Run a small first batch and read every transcript before you widen the audience.

When should an AI voice agent hand a call to a human?

Hand off on disputes, fraud claims, hardship, repeated failed steps, or any request for a person. Gartner found that 87% of 3,566 customers say a company using GenAI for service must offer a way to reach a human. Make sure someone picks up each handover.

How do you measure whether voice agent calls improve activation and retention?

Give each script one metric, such as KYC completion in 24 hours. Compare called customers with a similar group you did not call. Count only calls where the customer spoke.

Begin where customers leave

This week, pull last month's unfinished sign-ups and choose the script that fits them. Talk to the agent in your browser before any customer hears it.

Book a demo of Insight7