Customer

UKRGASBANK is a state-owned bank and one of the five largest banks in Ukraine by asset size (approximately USD 4.5 billion). It has been identified as a systemically important bank by the National Bank of Ukraine. For more than 20 years, the bank has served millions of clients through 211 branches across the country and employs over 3,800 staff members. UKRGASBANK positions itself as a leader in green banking, actively financing energy-efficiency and renewable-energy projects and playing a key role in the restoration of Ukraine’s energy infrastructure.

Challenge

As part of its long-term AI implementation strategy, UKRGASBANK engaged EVE.Calls to support AI adoption across customer communications, starting with the automation of client notifications related to upcoming and missed payments.

The automation was built on the bank’s existing internal policies and legally approved notification frameworks.

Ethical Debt Collection Communication

EVE.Calls provides a software platform for automating customer communications for banks and other organizations. One application of the platform is the delivery of controlled, ethical reminders about overdue payments.

In contrast to traditional collection practices, EVE.Calls replaces aggressive calls with company-approved, structured communication scenarios. These scenarios clearly explain the situation to the customer in a calm manner and offer constructive resolution options. For customers, this approach improves service quality and reduces stress, while for the companies, it helps preserve trust and long-term relationships. EVE.Calls is not involved in debt enforcement or “collection pressure,” ethical AI use is a core principle of our work.

Debt Collection Solution for the Bank 

The AI Voice Agent was created within the EVE.Calls SaaS platform using UKRGASBANK’s existing conversational scripts and in full alignment with the bank’s internal policies and regulatory requirements. The solution combines structured, rule-based logic with controlled generative AI elements, enabling compliant, ethical, and personalized client communication while fully adhering to financial-sector standards. The AI Voice Agent conducts conversations with an average duration of approximately 3 minutes, It:

• greets the customer by name and confirms identity,
• states the overdue amount,
• asks when the customer will pay,
• records their repayment plan,
• tailors answers to the customer’s situation,
• detects financial difficulties or disputes,
• redirects complex cases to a live manager when required or upon request.

The AI agent includes 95 tailored responses, allowing it to correctly interpret hesitant, unclear, or emotional reactions while maintaining respectful communication.

AI debt collection Conversation Example 

Transcript:

- Hello.

- Hello. 

- Good afternoon. My name is Victoria. I’m calling from UKRGASBANK. Please note that our conversation is being recorded. Olga Romanivna?

- Yes.

- Nice to meet you. You currently have an overdue balance, 7 days. The total amount is 38 USD. When are you planning to pay off the debt?

- Today

- Okay. Can you pay the full amount?

- Yes.

- What happened? Why have you violated the terms of the loan agreement?

- I simply didn’t receive the SMS, so I wasn’t aware. 

- The information is sent via SMS and is also available in the mobile app. If you are not receiving messages, please contact a branch or the Contact Center to update your contact details.

- Okay, thank you.

- Great. We are waiting for the payment. We recommend paying through the EcoBank app.

Advantages of the AI Voice Agent

• Significant reduction in call center load: 58 hours of human labor saved per 3,000 calls.
• High intent-capture accuracy: 99% accuracy in capturing customer payment intent, including payment amount and payment date.
• Regulatory compliance: TCPA, GDPR, and SOC 2 Type II compliance, ensuring consistent and reliable financial communications.
• Faster outreach: Zero days of delay in customer outreach, reducing the debt resolution cycle.
• In accordance with UKRGASBANK’s policy, calls that required special handling were transferred to live agents at the bank’s request. These cases represented 16% of all calls, while the remaining 84% were fully handled by the AI Voice Agent without human intervention.

Results of Debt Collection with Software

To evaluate the effectiveness of the Conversational AI communication in real banking processes, EVE.Calls applied its proprietary evaluation methodology. During the pilot phase, EVE.Calls reviewed and analyzed all call recordings with the AI Voice Agent. After the pilot results were validated, the bank independently evaluated the solution (commercial performance metrics related to debt collection were measured separately by the bank and are considered confidential) and approved further implementation. 

Following this approval, the AI Voice Agent was deployed in live operations, and from approximately 3,000 outbound calls processed across multiple successful campaigns, totaling more than 3,500 minutes of conversations, a representative 10% sample was selected for in-depth qualitative analysis. Using our methodology, the selected calls were assessed across various dimensions: logical consistency of the dialogue, relevance of responses, response latency, correctness of personal data recognition provided by the client, and tone of communication. Based on this analysis, the AI Voice Agent demonstrated over 89% accuracy in managing conversations in line with the approved scripts, even while the project remained in active development.

In the remaining cases (11%), minor non-critical issues were observed, such as brief response delays or occasional conversational imperfections. These did not result in incorrect information, policy violations, data exposure, or any negative impact on clients. All such issues are being addressed through ongoing system improvements.