Incomplete KYC is an identity check a customer starts and never finishes. It often shows up weeks later as a support ticket. Use automated outbound calls to reach the customer soon after they stall, instead of waiting for them to write in.
Trace each ticket back to its KYC gap
Start by proving the link in your own data. Pull the last 90 days of tickets and tag every one that mentions a status check, "can't withdraw", "account limited" or "where is my loan". Then match each ticket's user ID to that user's KYC status in your CRM or KYC provider. Count how many of the tagged users never finished verification.
The chain usually runs in four steps:
- The customer signs up and starts KYC, then stops.
- The account stays limited or blocked.
- The first transaction or funding attempt fails.
- The customer writes in or calls, and an agent spends a ticket explaining a step the customer could have finished earlier.
The limits in step 2 are by design. The FATF's digital identity guidance covers Recommendation 10, which requires customer due diligence, including identifying the customer and verifying who they are. So an unfinished check becomes a block, and the block becomes a ticket.
Tiered KYC makes this louder. Each blocked action tends to create a "why can't I do X" ticket. Add the customer's KYC tier as a field on every ticket, so you can see which limit triggers the most contacts.
At the end of this step you should have one number: the share of KYC-related tickets that come from users with an unfinished check.
Find where KYC stalls and why
Map your funnel stage by stage, then count the users lost at each one. A typical path is: signed up, ID started, document uploaded, selfie or liveness check, verified, funded. Look first at the biggest gap between two stages.
Next, find out why people stop. Common reasons:
- A blurry photo or a document that does not match the form.
- An app crash.
- No time right now.
- Distrust of the request.
- Unclear steps.
Sort these into two groups. A call can fix the human reasons. It can coach a customer through a clear photo, explain why the check exists, or book a better moment. A call cannot fix a crash or a liveness check that keeps failing. Send those to the product team.
Then split customer-caused drop-off from product-caused drop-off. Track small win steps (app download, KYC to tier one, loan started) apart from full conversion. If most users fail at the same screen, the product is the cause. If they stop after a clear start, the customer is.
Finally, rank segments by how many users sit in each and how many tickets each creates. Call the segment with the most tickets per stalled user first.
Trigger a call soon after the stall
Fire the call from an event, not from a list. Your app or CRM should send a signal when a KYC step sits unfinished. The campaign then waits a set window before it dials. Same day or the next day works well, because the customer still remembers the app. Add a stop rule: if the customer finishes the step during the wait, cancel the call.
Put a softer channel first. Send an SMS or push message at the stall, then let the voice call follow. The customer has already seen the reminder, so the call feels expected.
Build the campaign with a tool that gives you these controls. In Insight7 they are in beta:
- Goal and audience: one goal, such as "finish the ID upload", and one segment, such as users stalled at the selfie step.
- Contacts: a CSV upload or names and numbers typed in.
- Do-not-contact: a list that is honoured when a campaign is copied, so people added since the original are left out of the copy.
- Rules: a campaign type, consent marking, calling hours, retry rules and a do-not-call list.
- Test: talk to the agent in the browser with no phone number needed, then place a test call to your own phone.
Check the rules for each region you call into:
- US: the FCC's 2024 ruling puts AI voices under the TCPA, so callers need prior express consent.
- EU: Article 50 of the AI Act says an AI that talks directly with people must make them aware it is AI. It has applied since 2 Aug 2026.
- Everywhere else: check local telemarketing and data rules.
Script the call to finish KYC
Keep the call short and give it one job: get the missing step done. Use this order:
- Open with who and why. "I'm an AI assistant calling from [brand] about your account setup."
- Explain the check in one sentence. It is needed before the account can send or receive money.
- Name the one missing step. For example, "your selfie check is not finished."
- Answer common questions. Usually how long it takes, why it is needed and whether the call is real.
- Offer help. Resend the link or book a better time.
Buyers fear AI that sounds scripted or robotic, so write lines the way a person would say them.
Expect "How did you get my number?" If they gave it at sign-up and consent is recorded, say so. Never ask for an ID number, OTP or PIN on the call. Send the link only through the official app, SMS or email.
Plan the handover before launch. Buyers ask whether it is polite, what happens with a disputed amount, and who chases a handover nobody picked up. Gartner's 2026 survey found 87% of customers say a human option is essential, so offer one on every call.
End every call with one outcome: will complete now, needs help, refuses or wrong number. Each outcome routes somewhere, and a ticket carries the reason.
Measure ticket drop, not just calls
Set your metrics before launch, because measurement is where AI service projects fail. Gartner's review of 432 AI customer-service use cases found 42% had unclear ROI, meaning leaders could not say what value they got.
Track three groups:
- Outcome: KYC completion within 7 days of the stall, hours from stall to completion, KYC-related tickets per 1,000 new users, and first-funding rate.
- Call: reach rate (a human answered), connect-to-completion rate, retries per contact, and handover rate.
- Guardrail: opt-outs, complaints about the calls, and escalations.
If outcomes improve while guardrails worsen, you are buying completions with annoyance.
In Insight7, outbound campaigns are in beta. A dashboard shows calls in progress and lets a reviewer read each transcript, and yes/no answers and call outcomes are pulled from the transcript. Calls where only the AI spoke are not counted as reached, and a carrier's temporary refusal does not use up a retry, so your reach rate stays honest. Campaigns also carry consent marking, calling hours and retry rules.
Prove cause with a control group. Hold out a slice of stalled users who get no call, or only the SMS. Compare completion and tickets between the groups. Without that, you cannot tell a call-driven completion from one that would have happened anyway. This also helps when SMS and voice run in parallel, since each needs its own comparison group.
To size the saving, use this formula: extra completions × tickets avoided per completed user × cost per ticket, minus the cost of the calls. Every input is an assumption to replace with your own data. Here is an illustration with made-up inputs, not real results. Say completion rises by 150 users over the control group. Say each completed user avoids 0.5 tickets, and a ticket costs $8. That is 75 tickets avoided, or $600. If the calls cost $400, the net saving is $200 before any funding revenue.
Review weekly. Read the transcripts of failed calls, because they show which script line loses people.
Related reading: How to Use AI Voice Agents to Recover Customers Who Drop Off During Onboarding, AI Outbound Call Script Templates for Abandoned Signups and Incomplete Applications and How to Increase Account Funding Rates for Fintech and Investment Apps With AI Outbound Calls.
Questions people also ask
What happens if KYC is not completed?
Usually the account stays limited, and the customer cannot do things like fund, withdraw or borrow until the check is done. Rules differ by provider and region, so check yours. Each blocked action tends to become a support contact.
What does incomplete KYC mean?
A required part of the identity check is missing. The customer may have skipped a selfie, uploaded a blurry document, or had a submission rejected or expired. The status is not "verified."
How long should you wait before calling a customer who stalled on KYC?
Wait a few hours to one day, after the SMS or push reminder has gone out. That gives people time to finish alone. Then test call windows, since answer rates differ by time of day.
Can an AI voice agent call customers about KYC?
Yes, where local rules on consent and AI disclosure are met. Get and record consent, state that the caller is an AI, and respect do-not-contact lists.
How many times should you retry a KYC reminder call?
Treat three to four attempts, spaced across different times of day, as an assumption to test against your own reach data. Many people answer on a later try. Stop at once on an opt-out or a completed step.
Should a KYC reminder call ask for ID details?
No. The call should only name the missing step and point to the official app or link. Asking for numbers by phone looks like phishing and teaches customers a risky habit.
How do you measure whether KYC calls reduce support tickets?
Count KYC-tagged tickets per 1,000 new users over a fixed window, such as 30 days after the stall. Compare the called group with a held-out group, using the same tagging rules for both.
Fix the stall before it becomes a ticket
A stalled check costs you a call now or a ticket later. Set the trigger, give the call one job, and measure tickets against a control group. Call at the stall, not after the complaint.

