Boosting productivity with GenAI Claire Assistant

20 %

of onboarded employees use Claire on regular bases

38 000+

conversations started with Claire over the course of 7 months

About the Client

Erste Digital is the IT competence center of Erste Group, responsible for delivering digital solutions and managing core banking systems across all of the Group’s markets. Headquartered in Austria and employing over 2,000 people, it is one of the largest IT companies in Europe. As part of Erste Group, a major financial services provider in Central and Eastern Europe with 48,000 employees, 11 billion euros in revenue, and 16.6 million customers, Erste Digital plays a critical role in driving innovation and operational efficiency.

In collaboration with Erste Bank Österreich, which serves 4.3 million customers and employs more than 16,000 people, Erste Digital is leading the transformation of internal processes. This joint effort focuses on leveraging AI-driven solutions to enhance employee productivity and streamline banking operations across the Group’s seven core markets.

The Challenge

Employees at Erste Bank Oesterreich are searching for all kinds of instructions for their work daily. They look for information about custom deposit conditions, exceptions, processes for specific money transfers, or differences between loan collaterals. There are hundreds of documents and process books, which are difficult to navigate through and time-consuming to locate the appropriate instruction for each specific case and customer. With thousands of employees, even a small improvement would positively impact the bank’s operational efficiency, meeting the rising demand for seamless, effective internal support.

The Solution

To boost efficiency of employees and reduce the need for manual search within piles of documents, Erste Bank decided to develop a chatbot that efficiently identifies correct documents and provides the user with relevant paragraphs incl. links and explanations.

As the technology of choice was selected Databricks, a recently implemented AI platform across Erste Group, which offers variety of GenAI functionalities (such as Vector search, REST API and Unity Catalog). For this kind of use case, a Retrieval Augmented Generation (RAG) pattern is ideal. This approach transforms original documents into vector embeddings and let the user search in it using spoken language and Large Language Model (in this case GPT 32k).

The source data - documents and process books - are conveniently stored and ingested via secured Sharepoint and Azure storage, while the solution orchestration backend is hosted on Erste on-premise server. This setup secures high control over the solution and also allows seamless integration with existing employee assistants.

Finally, users are encouraged to provide their feedback, using the thumb-up and thumb-down buttons. This input ensures ongoing stream of feedback used for further improvements and performance monitoring over time