Performance Technologies S.A.
Contact: Dimitra Kontou / Michalis Michail
Country: Greece
Award Category: Innovative Problem Solver
Tell us about the business problem you were trying to solve.
The Analytics Center of Excellence Department of the National Bank of Greece, in collaboration with strategic partner Performance Technologies SA, has implemented an innovative Digital Transformation project for the development and application of Customer Service using Generative AI & LLMs technologies, empowered by Real Time Decisioning, for the immediate and automated categorization of the Bank's customer requests.
National Bank of Greece was facing a huge workload derived from unstructured data of customer service requests, that:
a) required a significant amount of back end human agents to analyze, process and respond,
b) each human agent needed an important amount of time to be trained, efficient and productive,
c) customer requests most of the times, did not include important other information for the customer profile, customer value, products etc., so the analysis of its request needed much time to collect and complete the required relevant information
d) this process was high critical to full fill in a timely manner and many times included high risk, as these requests maybe related to fraud cases, or loan approval processes or other business critical issues, for which immediate action was required.
The project's goal was to improve the Business Process Model of customer service through the analysis of texts and metadata related to hundreds of thousands of requests submitted annually on the Bank's website. Specifically:
1. Request Analysis and Categorization: Using Machine Learning and Generative AI techniques for categorizing requests, analyzing sentiment, and assessing operational criticality.
2. Real Time Decisioning: Enriching the customer request with additional customer profiling and customer behavior data, applying business logic to enhance and automate the decision process for the importance and timeline for each response.
3. Suggested Responses: Creating suggested response guidelines for the Bank's Agent and generating personalized customer facing response texts.
4. Routing Results: Routing the analysis results in real time to the Bank's central management system.
5. Immediate Automated Analysis for Faster Request Handling: Real-time analysis of each request, directly impacting the reduction of request handling time.
6. Improving Customer Experience and Satisfaction: Through immediate, focused, and optimal utilization of the Bank's communication rules and policies.
The challenges the project team faced are:
1. Analysis Accuracy: The need for high accuracy in the analysis and categorization of requests to ensure proper customer service.
2. Adaptation to Greek Data: The development and adaptation of Artificial Intelligence models for the Greek language and the specific characteristics of customer requests in Greece.
3. System Integration: The integration of analysis results with the bank's existing systems in real-time.
What SAS products did you use and how did you use them?
Our strategy for implementing the project included analyzing the Bank's requirements and selecting the appropriate technology to meet these needs. Initially, we conducted a pilot application using the SAS VIYA platform on Azure, SAS Intelligent Decisioning and Azure OpenAI service, where we utilized Large Language Models (LLM) and Generative AI to respond to customer requests and suggest response texts. Additionally, alongside the use of SAS Text Analytics, we identified cases of requests and texts where the use of traditional NLP techniques yields optimal results. We reviewed the results of the pilot project, decided on the optimal design of the solution, and proceeded with the next steps for its full implementation.
Our creative idea was based on leveraging the most advanced AI technologies for automating and optimizing customer service. The implementation process included analyzing and designing new processes, supporting the Bank in implementing these processes, and integrating the results with the Bank's tool that manages cases/requests. With this approach, we managed to create a solution that not only improves the customer experience but also exponentially increases the efficiency of service.
More specifically the SW tools within architectural blueprint are:
- SAS VIYA Intelligent Decisioning: for the real time processing of incoming Customer Service requests, the enrichment with other Bank sources (DWH, CRM etc.), real time decisioning logic, orchestration with Bank systems (Content Management etc.) and real time integration with external SW (OpeAI on Azure).
- SAS VIYA Text Analytics: for parsing of unstructured data for sentiment and topics/categories.
- OpenAI ChatGPT 4o: for classification on categories, subcategories (Prompt agents), and RAG(Retrieval Augmented Generation) model for automated recommandation of answering templates to customer requests.
- OpenAI ChatGPT 4o mini: for sentiment classification (negative sentiment or not)
- OpenAI Text Embedding 3 Large: for RAG (Retrieval Augmented Generation) so to drill down and analyze existing Bank knowledge base of answering templates and propose to the human agent a customer facing text (prepared again with ChatGPT 4o) for a response to a request.
- LangChain Python Library: for RAG (Retrieval Augmented Generation), it acts as an orchestrator between the GenAI model and the Embedding model.
The high level process flow is as below:
- Customer submits a request on the Bank’s web site and within the Bank’s Content Management Tool
- With Rest API approach this request is received from SAS VIYA Intelligent Decisioning (ID)
- SAS VIYA ID enriches the request with lookups/query on other Bank systems (DWH, CRM etc.), retrieves additional customer profile information (customer segment, products etc.),
- SAS VIYA ID integrates with Azure OpenAI via a specific Bank bridge, to execute a series of LLM Agents (classification, response etc.),
- SA VIYA ID applies business decision logic/algorithms,
- SAS VIYA ID provides the automated response to the REST API call back to the Bank’s Content Management Tool
- Human agent receives in real time the GenAI driven response (enriched, classifies, prioritized, response text etc.)
- The Human agent utilizes this automated input and respond to customer instantly, or enhances/modifies the proposed response or escalates this response to 2nd level/if needed.
What were the results or outcomes?
The effectiveness of our project in categorizing customer requests at the National Bank of Greece was excellent and fully met the goals we had set. The results we achieved are measurable and demonstrate the effectiveness of our solution. Specifically, we achieved automated categorization of requests with an accuracy rate of 90%.
The use of Generative AI and Machine Learning technologies allowed us to analyze the sentiment of these requests and use it in the management/response process.
The optimization of the prioritization process leads to a reduction in response time for a request. We estimate that this improvement, along with the creation of suggested response guidelines for the Bank's agent, improves the resolution time of all requests by 25%.
At the same time, we significantly increase the satisfaction level of our customers with the overall management of their requests. Our solution incorporates innovative AI technologies that improve the efficiency and accuracy of customer service.
The benefits for the company and its customers are manifold. From a process perspective, our solution automates and optimizes the analysis and categorization of requests, reducing response time and improving service quality. The bank's staff benefits from the automation of processes, as they can focus on more strategic and creative tasks.
The financial benefits for the business are also significant. Improving the efficiency and accuracy of customer service leads to cost reduction and increased customer satisfaction, which in turn leads to increased bank revenue and customer loyalty.
Finally, the social impact of our solution is positive, as it improves the customer experience and enhances their trust in the Bank and the financial system in general.
Why is this approach innovative?
In the new era of AI and GenAI it is paramount the efficient management and orchestration in real time of the many AI specialized models and agents, together with company wide specific data sources, applications, internal tools, operating procedures, policies and other.
This key requirement that mandates the utilization of powerful and user friendly real time decisioning and orchestration software, like SAS VIYA platform and SAS Intelligent Decisioning.
Within the chosen solution we included Generative AI and Machine Learning technologies, specifically a best-of-breed approach was selected with the SAS VIYA software and the Microsoft Azure OpenAI Service. This solution incorporates specialized Artificial Intelligence models for the Greek language, allowing for high-accuracy analysis and categorization of requests.
The innovative elements of the solution include the use of NLP (Neuro Linguistic Programming) techniques, LLM (Large Language Model) such as GPT-4.0, etc., for analyzing the sentiment and criticality of each request, as well as RAG (Retrieval Augmented Generation) techniques for creating suggested guidelines and response texts. Additionally, the solution includes the real-time integration of analysis results (Real Time Intelligent Decisioning) with the bank's management system. Using SAS Viya, information was enhanced with metadata from other bank exercises, such as customer information, which led to the optimization of the request prioritization process.
The solution is scalable and is planed to include many new use cases of real time decisioning empowered with AI and GenAI for the complete customer journey, service management, marketing and sales communications.
Based on the above, we believe that our nomination demonstrates the effectiveness and innovation of our solution while offering significant benefits for the company, customers, financial sector and to the society.
