Large Language Models for Data-Driven Customer Relationship Management: Transforming Unstructured Data into Business Intelligence
Sundstedt, Vilhelm (2024)
2024
Julkaisu on tekijänoikeussäännösten alainen. Teosta voi lukea ja tulostaa henkilökohtaista käyttöä varten. Käyttö kaupallisiin tarkoituksiin on kielletty.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi-fe2024091170617
https://urn.fi/URN:NBN:fi-fe2024091170617
Tiivistelmä
The research conducted in this thesis highlights the transformative potential of LLMs for analyzing and creating BI from unstructured data, a crucial development in the era of big data. The use case of Mirka Ltd. exemplifies how LLMs can address persistent challenges in extracting value from the vast amounts of unstructured data typically underutilized by businesses. Through the development and evaluation of an LLM-based system, this study successfully demonstrates how unstructured customer visit notes can be converted into actionable insights for subsidiary Managing Directors, contributing to improved management and enhanced sales management reporting.
From a design research perspective, the research highlights the importance of working systematically when conducting LLM-based analysis, especially when dealing with large volumes of unstructured and company-specific data. The study shows that with the right approach, the inherent limitations of current LLMs can be mitigated and overcome, enabling the creation of a scalable and accurate system for unstructured data analysis and BI reporting. This enhances the value derived from previously neglected data sources and lays an adaptable foundation for continuous improvements as generative AI advances.
From a financial and behavioural perspective, the LLM-based system offers substantial benefits. By automating the extraction of key insights from the unstructured visit notes, the Managing Directors are equipped with relevant and concise information when producing their sales management reports. Additionally, the system enhances data-driven decision making, potentially leading to significant cost savings and increased revenue. Since the analysis system is relatively cost-effective, these benefits significantly outweigh the low cost of deployment.
Looking ahead, generative AI and LLMs present significant opportunities in BDA and BI. The LLM-based analysis developed in this thesis could be extended to Sector Managers and other stakeholders who would benefit from the insights extracted from visit note data. Additionally, there is potential to apply LLMs to other types of analysis, such as topic modelling and sentiment analysis. As LLM technology continues to evolve, integrating more advanced and efficient models can enhance performance and reduce costs of these analyses even further. Developments in generative AI could one day lead to reliable descriptive, predictive, and even prescriptive analytics, with the potential for direct and accurate question-answering over data.
In conclusion, this thesis demonstrates how LLMs can revolutionize the way businesses extract value from unstructured data, leading to improved BI and advantageous business outcomes. By employing a systematic design research approach when utilizing LLM-based analysis, a balance between accuracy, scalability, and cost efficiency can be achieved. As AI continues to advance, the ability to leverage LLMs for unstructured data analysis will be increasingly important for organizations aiming to maintain a competitive edge in a data-driven world.
From a design research perspective, the research highlights the importance of working systematically when conducting LLM-based analysis, especially when dealing with large volumes of unstructured and company-specific data. The study shows that with the right approach, the inherent limitations of current LLMs can be mitigated and overcome, enabling the creation of a scalable and accurate system for unstructured data analysis and BI reporting. This enhances the value derived from previously neglected data sources and lays an adaptable foundation for continuous improvements as generative AI advances.
From a financial and behavioural perspective, the LLM-based system offers substantial benefits. By automating the extraction of key insights from the unstructured visit notes, the Managing Directors are equipped with relevant and concise information when producing their sales management reports. Additionally, the system enhances data-driven decision making, potentially leading to significant cost savings and increased revenue. Since the analysis system is relatively cost-effective, these benefits significantly outweigh the low cost of deployment.
Looking ahead, generative AI and LLMs present significant opportunities in BDA and BI. The LLM-based analysis developed in this thesis could be extended to Sector Managers and other stakeholders who would benefit from the insights extracted from visit note data. Additionally, there is potential to apply LLMs to other types of analysis, such as topic modelling and sentiment analysis. As LLM technology continues to evolve, integrating more advanced and efficient models can enhance performance and reduce costs of these analyses even further. Developments in generative AI could one day lead to reliable descriptive, predictive, and even prescriptive analytics, with the potential for direct and accurate question-answering over data.
In conclusion, this thesis demonstrates how LLMs can revolutionize the way businesses extract value from unstructured data, leading to improved BI and advantageous business outcomes. By employing a systematic design research approach when utilizing LLM-based analysis, a balance between accuracy, scalability, and cost efficiency can be achieved. As AI continues to advance, the ability to leverage LLMs for unstructured data analysis will be increasingly important for organizations aiming to maintain a competitive edge in a data-driven world.