The Client
Credem, founded in 1910 and later becoming Credito Emiliano SpA, is one of the leading private banks in Italy. Known for its technological innovation and personalized services for private and corporate clients, it introduced WELLBANKING in 2020, emphasizing financial and social well-being. It also stands out for its multi-channel distribution network, combining traditional branches with online services.
Services provided
OBJECTIVE
The goal is to improve the customer experience on Credem's website by effectively managing the access to main information about products and increasing the conversion rate towards purchase, additionally try to solve customer issues or redirect them to the customer care center the the thousands of daily requests, which are often repetitive and cause long waiting times.
Solution
Emily is the virtual agent, based on Artificial Intelligence, created by Credem and AND EMILI to assist anyone needing information on banking services. Emily interacts with users through text and voice and uses an advanced cognitive architecture that incorporates state-of-the-art Large Language Models along with other technologies in the field of Natural Language Processing (NLP). This architecture integrates various NLP techniques, such as entity extraction, tagging, and embedding, to provide more accurate and relevant answers.
Emily is available on the Credem.it website and WhatsApp and can connect to various external services through APIs, responding promptly to user questions 24/7. Her knowledge base is dynamically updated, ensuring that Emily is always aware of changes and constantly updated with the latest information. This enables her to assist users in common procedures like password recovery, provide personalized advice, suggest types of current accounts, and guide them in finding specific information on the site. Emily is capable of understanding the context of less clear requests, improving interaction through follow-up questions.
She is designed to enhance the quality of her responses over time, deriving value from interactions with users.
Emily's agency skills are not limited to responding to linear questions; she acts as a thinking entity, guided by a hierarchical structure of protocols and processes. This allows her to reason and choose the most appropriate tools to complete her task. For example, she can perform calculations for a hypothetical loan a customer might request, geolocate Credem ATMs and offices, or navigate the website as a human would to find the most relevant document before responding carefully to the user.
Security
Emily is developed considering the highest requirements of cybersecurity and personal data protection.
The solution is based on a serverless microservices architecture entirely managed by Amazon AWS infrastructure, which guarantees both security and scalability. Amazon Aurora PostgreSQL, AWS Lambda, and Amazon ECS Fargate are some of the services used by Emily's infrastructure and integrate with OpenAI, Azure OpenAI, and Llama 2 language models.
The solution uses AWS's server architecture, employs IAM (Identity and Access Management) to manage access to AWS resources securely, and encrypts sensitive data in transit and at rest using services like AWS Key Management Service. Regarding OpenAI, data is anonymized before sending according to GDPR requirements.
