AI-Powered Banking Digital Assistant for a UK Commercial Bank

Project background

The banking industry in the UK is witnessing a profound transformation and shifting to the power of advanced AI technologies and solutions, like financial digital assistants, to automate routine tasks, optimize investment strategies, and elevate the overall customer experience. As statistics in the UK’s banking sector indicate a growing reliance on technology and AI-driven solutions, the benefits for businesses like our client’s bank are evident. Such a software solution will not only enhance operational performance but also help attract and retain customers in an increasingly competitive landscape.

Our client, a forward-thinking financial institution in the UK, is committed to staying at the forefront of innovation and technology in the rapidly evolving fintech landscape. With a strong emphasis on security, efficiency, and cutting-edge solutions, the bank has recognized the need for a custom transformative software solution for banks to bolster its position in the market. In this context, they have sought to partner with a professional team possessing extensive expertise in banking software solutions development. 

This project is multifaceted and encompasses several critical components. One of the key objectives is to enhance the system’s performance and scalability to meet the growing demands of users. This includes reducing latency and ensuring that the infrastructure can handle increased workloads seamlessly. 

Furthermore, the project focuses on developing a modern banking infrastructure, incorporating cutting-edge AI and machine learning technologies, including a virtual financial assistant. This assistant will streamline customer interactions and improve efficiency while delivering an exceptional customer experience.A notable challenge the project addresses is extracting critical information from a large volume of unstructured data contained within bank contracts. By doing so, our solution will streamline data processing and analysis, allowing for more informed decision-making. Additionally, developing a conversational AI-powered banking chatbot will provide round-the-clock assistance to customers, offering access to open banking information and fulfilling customer requests while ensuring data security. Moreover, our project considers concerns and on-premise service infrastructure with a secure and smart information search solution utilizing Natural Language Processing (NLP) and Retrieval Augmented Generation (RAG) technologies.

Business needs

The client’s bank prioritizes innovation, security, and efficiency to provide a competitive advantage in the fintech landscape. This software solution for banks aims to establish a cutting-edge commercial bank in the UK that leverages advanced AI solutions to enhance security and customer experience, automate routine tasks, and optimize investment strategies. Our client was looking for a professional team with vast experience in banking software solutions development. 

– Performance and Scalability: Enhance system performance, reduce latency, and ensure scalability to meet increasing user demands.

– Modern Banking Infrastructure: Create a modern banking infrastructure with virtual financial assistant by integrating AI and machine learning technologies.

– Key Information Extraction from the Bank Contracts: processing a large amount of unstructured data from bank contacts.

– Conversational AI Chatbot: Develop a conversational AI-powered banking chatbot capable of assisting customers 24/7 with open bank information and requests while maintaining data security.

– Secure and Smart Information Search for Internal Use: NLP Solution for Retrieval Augmented Generation (RAG) taking into account security concerns and on-premise service infrastructure.

– Improve User Engagement and Attraction: Enhance user engagement via banking AI chatbot development, deliver added value to existing users, and differentiate services from competitors.

– Rigorous Testing: Conduct thorough testing to validate the accuracy and reliability of AI solutions.

– Continuous Monitoring and Updates: Implement continuous monitoring and updates during finance software solution development to ensure ongoing improvement in system performance and functionality.

Product features

1. Processing of banking documentation in financial digital assistant:

– Optimized and automated work with numerous bank contracts for the purchase of financial instruments.

– Extraction of the unstructured contract data: 

– Counterparty Information:

Legal Entity Name: The full legal name of the contracting party or parties involved in the agreement.

Contact Information: Contact details, including addresses, phone numbers, and email addresses, for communication between parties.

– Contract Agent Information:

Agent Name: The name of the authorized representative or agent acting on behalf of one of the parties.

Agent Contact Information: Contact details for the agent, including their role and authority in the contract.

– Agreement Conditions:

Contract Title: A descriptive title or name for the contract which helps identify the nature of the agreement.

Purpose: A clear statement outlining the purpose or objective of the contract, such as investment, purchase, sale, or other financial activities.

Terms and Conditions: Detailed terms and conditions that specify the rights, responsibilities, and obligations of each party. 

– Duration and Termination:

Contract Start Date: The date on which the contract becomes effective.

Contract End Date: If applicable, the date on which the contract expires or is terminated.

Termination Conditions: Specific conditions or events that allow either party to terminate the contract before its scheduled end date.

– Payment and Compensation:

Payment Amount: The total amount to be paid under the contract, including any applicable interest or fees.

Payment Schedule: A schedule outlining when payments are due, including due dates and payment methods.

2. Investment Services:

– Personalized investment recommendations via banking digital assistant;

– Performance tracking and reporting for investments.

3. Access Channels:

– Mobile banking app for on-the-go access.

– User-friendly web portal for desktop and tablet users.

– Virtual financial assistant, powered by GPT model.

4. Customer Support:

– 24/7 customer support with instant responses about general issues of the bank.

– Personalized banking digital assistant for general inquiries and advice.

– Seamless integration with other bank services.

– Smart information filtering.

Solution

During the development process, our development teams did the following:

1. AI-Powered Chatbot: developing question answering-system for banking

– Design and further banking AI chatbot development using the GPT model to provide personalized customer support;

– Enable the chatbot to handle inquiries related to general bank information and queries like ‘How to open/close an account’, ‘What documents are needed for procedures’, etc.;

– Ensure the chatbot can escalate personalized queries involving sensitive client data to human agents.

2. Tasks Automation:

– Identify routine and repetitive tasks within the bank’s operations.

– Develop AI-driven automation for banking to streamline these processes.

– Examples include investment document processing and report generation.

3. Smart Search and Filtering for Internal Use: 

– On premise banking software solution development to process secure bank information based on the Large Language Model to avoid possible data leaks in the cloud

– Retrieving, generating, and filtering information while prioritizing security and utilizing on-premise service infrastructure.

4. Custom Entity Extraction: 

Custom entity extraction within a banking digital assistant involves tailoring natural language processing (NLP) models to recognize and extract specific entities and information, such as:

– Financial Instruments: Identify and extract information about financial instruments such as stocks, bonds, derivatives, options, and mutual funds.

– Transaction Details: Extract details about financial transactions, including transaction amounts, dates, parties involved, and transaction types (e.g., deposits, withdrawals, transfers).

– Banking Products: Recognize and extract information about banking products and services, including investment, savings accounts, loans, credit cards, etc.

– Bank Names and Branches: Identify the names of banks and specific branch locations mentioned in the text.

– Account Numbers: Extract account numbers or identifiers related to bank accounts.

Initially, before banking chatbot development, our expert team executed meticulous performance testing and proposed the strategic shift from Spray HTTP to Akka HTTP to optimize server application productivity. 

– Technology stack upgrade: Implement a strategic shift from Spray HTTP to Akka HTTP and adopt the (Twitter) Finagle HTTP client for improved server application productivity and performance.

– Performance optimization: Utilize performance testing and monitoring tools (opentracing, Zipkin, Grafana Dashboards) to enhance system performance, identify bottlenecks, and ensure optimal server response times.

– Data serialization improvement: Migrate from Kryo to Protobuf, leveraging Scala Macros for proto file generation, to optimize data serialization/deserialization and enhance development productivity.

– New customer attraction: Strategically employ newsfeed aggregation as a supplementary information service within the bank’s systems to attract new customers and improve user engagement

Our technology stack

• Azure Open AI
• Azure Form Recognizer
• OpenCV
• finBERT
• Tesseract for OCR
• pdfminer
• Python
• RegExp
• LSTM
• Keras

Programming Languages and Frameworks:

• Scala macros
• Scalameta
• MacWire
• scalaxb/soap
• spray (json, HTTP)
• Finagle HTTP
• SBT
• Gatling

Client values