AI voice agent for overdue payments
Duecall calls the customer, learns why the payment is late and writes the agreed date and amount to your CRM.
Pilot: 2 weeksup to 500 customersfree
You play a customer with an overdue payment; the agent talks as on a real call. At the end you see the result card that goes to the CRM.
Sample data. Up to 90 seconds, no sign-up. The demo transcript is kept to improve quality — do not say real personal details.
In the browser — through your microphone, right now. On the phone — the agent calls the number you confirm with an SMS code. To talk to a person — use the request form below.
Try the live demo a little later, or request a pilot and we will show it ourselves.
Headphones sound better: the agent's voice does not leak into the microphone.
The agent calls you over a real phone line — with the voice and quality your customers will hear. The scenario is the one chosen above.
No phone line is connected right now: confirm your number and the agent will call later.
Every day there are more customers to call than operators can handle. Part of the portfolio is left untouched; reasons and promises get lost in notes.
Operators reach a fraction of the list in a day. The rest slips to tomorrow, then next week — while the delay grows.
The agent calls the whole list inside the calling window; operators keep only the hard cases.
Why the customer did not pay and when they will stays in an operator's note. Useless for reports, filters and control.
The reason as a category, the promise with a date and amount. The same structure from every call.
Calls are only allowed in a limited window. A break, a meeting, a sick day — each one is lost calls.
Parallel lines, no breaks; no calls outside the set window.
The pilot report comes with exactly these metrics — we do not promise any numbers in advance.
Your process does not change: the list from you, the call from us, the result back to you.
CSV/Excel, API or the admin panel: name, phone, product, amount due and days overdue. Who the agent speaks for and in what tone (the playbook) is agreed with you.
You geta call schedule and an approved playbook.
During working hours, from the agreed number. It checks identity, asks why, agrees the payment date and amount — in whichever language the customer speaks.
You geta reason, a promise and a record per customer.
Status, reason, agreed date and amount, requests, summary and transcript — by webhook, API or CSV, right after the call.
You geta ready queue for operators — only the hard cases.
The same fields from every call. Ready for filters, reports and the next step.
The agent works inside a strict script; where and how long data is kept is fixed in the contract; the result arrives in your system.
The list, results, transcripts and recordings — on our server in the cloud option, only in your infrastructure on-premise. The retention period is agreed with you.
External AI services are used for speech recognition, reply generation and voice synthesis — the audio and text of the conversation are sent to them. Which ones and on what terms is stated explicitly in the contract.
The demo transcript is kept to improve quality; in the phone demo your number is used for the SMS confirmation. Do not say real personal details.
However you hand over the list, that is how the result comes back.
Upload the file in the admin panel, columns are checked, the result comes back as a file or in the panel.
Send the list, read call status; after every call a signed JSON reaches your endpoint.
The customer is found by phone, the result is written to the contact or deal timeline; a message to a Telegram group.
Connected on top of the API/webhook together with your IT team.
In an on-premise project, built separately for the bank's internal system.
{
"event": "call.finished",
"call": {
"outcome": "promise_to_pay",
"identity": "confirmed",
"reason": "salary_delay",
"promises": [
{ "amount": 1850000, "date": "2026-10-15", "strength": "firm" }
],
"summary": "…",
"transcript": [ … ]
},
"requests": [ … ]
}Pricing is individual — by portfolio size, number of calls and deployment type. What it is made of is clear up front.
MFOs, instalment retail, telecom, utilities
Banks
The exact quote comes after the pilot: based on real minutes, reach and results.
Short and direct.
Uzbek and Russian. If the customer switches language mid-conversation, the agent switches too. It understands mixed (uz-ru) speech.
The list, results, transcripts and recordings are kept on our server in the cloud option and only on yours on-premise. External AI services are used for speech recognition, reply generation and voice synthesis: the audio and text of the conversation are sent to them. Which services, on what terms and for how long data is kept is fixed explicitly in the contract.
Yes. For banks the system runs on your servers: your own Asterisk/SIP trunk and number; the list and results never leave your perimeter. Options for the speech services are agreed in the project. The cloud option starts faster for MFOs and instalment retail.
A pilot via CSV/Excel — within a few days: upload the list, approve the playbook, calls begin. API/CRM integration and an on-premise project are planned separately.
No. A promise is the date and amount the customer named. The agent marks it firm or tentative; the actual payment is checked in your system, and on the promised day the agent makes a reminder call.
The agent does not argue or pressure. It listens, records the request and flags the conversation for an operator. Such calls are shown separately in the report.
No. The agent takes the repetitive calls: remind, learn the reason, get a promise. Operators handle disputed, complex and large cases — with the context prepared.
The simplest route is CSV/Excel in the admin panel. Then REST API and webhook: a JSON arrives after every call. Bitrix24, amoCRM and Telegram have a ready connection; 1C and core banking are per project.
Calls go out from the official number, in the configured working hours, with a limited number of attempts. One customer is not called many times a day, and a “do not call” request is respected at once. This lowers the risk of a spam flag but cannot rule it out entirely.
Onboarding and setup (one-off), the platform fee (monthly or an annual licence), conversation minutes and telephony. The exact figure depends on portfolio size and deployment type — we quote after the pilot, on real numbers.
Leave a short request: we get back within one business day and plan a free 2-week pilot.
Write on Telegram →Pilot: 2 weeks · up to 500 customers · free