Piloting AI for improved efficiency and decision-making in operations and customer support

Published: 23 October 2024

Improving customer support with AI

One of the main projects in the Cards business unit focuses on making customer support more efficient by automating customer verification. With about 750,000 customer service calls handled annually, even saving a few seconds per call can make a huge difference. The solution currently being developed is a voice bot using machine learning models to handle basic verification, allowing customer service agents to focus more on complex issues such as human interaction and problem solving.

"By automating the initial verification steps in a call, our customer service agents can focus on what truly matters. They can connect with the people and solve the actual problems that the customers reach out to us to get help with," says Marit Engh, Head of Customer Support & Dispute Services at Tietoevry Banking.

The project is in the proof of concept (PoC) phase, aiming for full integration into the support process by next year. If successful, the AI solution could reduce wait times and enhance customer experience by automating routine tasks, allowing more time for human interaction in addressing unique issues and customer inquiries.

Predictive AI for data analytics through dashboards

Internally, the Cards unit is building AI capabilities to improve data analytics and reporting. Using Microsoft's Power BI and other tools, the team is working on predictive analytics that can forecast trends in card usage.

The AI model is currently being tested based on three years of support data, covering more than one million cases. On this backdrop, the system can identify patterns and generate accurate predictions for future trends and changes in card usage. Right now, this particular AI project is in the testing phase, being used internally to support better decision-making. The goal is to offer more advanced reporting and prediction tools to banks going forward, helping them optimize their own operations.

Speeding up product development

The Cards team is also exploring how AI can streamline product development and testing by automating key parts of the process. Tools such as GitHub Copilot and internal AI-driven chatbots are being used to generate test cases and test data, reducing the need for manual work while allowing new products to be tested faster and more accurately.

In product development, AI is also planned to help with customer solution specifications, making it easier to customize products to meet different client needs. Instead of multiple workshops and lengthy back-and-forth interactions between different domain experts and teams, AI can help adjust parameters and move the process forward more quickly, speeding up the time it takes to draft and finalize a product proposal.

In addition, AI plays a crucial role during migrations, where extensive testing is necessary to ensure everything works correctly when transitioning to new systems. By automatically generating test data, AI helps make the process more efficient, ensuring the right data is used and reducing the time required to validate systems before they go live.

"These AI solutions help us streamline repetitive tasks such as creating test cases. In that way, we can run tests and bring new products to market faster and avoid time-consuming and sometimes tedious and repetitive tasks," says Marit Engh.

Always balancing AI and compliance, while avoiding typical AI-pitfalls

While AI offers many efficiency benefits, Tietoevry Banking is careful to ensure that all solutions comply with existing data privacy regulations, such as GDPR, and upcoming laws like the EU AI Act. The Cards unit has taken on a "privacy by design" approach, ensuring that AI is implemented in a way that respects data security and regulatory requirements. This is part of the framework that Tietoevry Banking as a whole is using when designing and developing new AI solutions.

The team is also mindful of typical challenges such as AI hallucinations, where the large language models (LLM) may generate inaccurate or misleading information. To manage this, they fine-tune AI models and use methods such as a document-specific training called retrieval augmented generation (RAG), to ensure the AI operates within the limits of pre-validated, reliable data.

With these initiatives, Tietoevry Banking's Cards business unit is cautiously advancing AI technologies to improve both customer-facing services and internal processes. The unit expects to see significant results from these efforts by 2025, with AI helping to improve efficiency, reduce manual and repetitive tasks, and support better decision-making.

Click here to learn more about Tietoevry Banking's card services and processing.