Comparative analysis of LLM capabilities and their implementation potential in business workflows

Are large language models in their current state capable of optimizing internal processes to speed up the workflow? What are the current technologies and tools available to achieve this goal of an artificial intelligent driven business?

Kevin Bernet, 2024

Art der Arbeit Bachelor Thesis
Auftraggebende Company XY
Betreuende Dozierende Jüngling, Stephan
Keywords Process optimization, large language model
Views: 7
Company XY is still growing and the processes have changed over the last few years. Therefore, they have never been drawn and analyzed systematically to see if there is any potential for optimization. For this reason, the author was commissioned to do so, and the following research questions have emerged: Are large language models capable of optimizing internal processes? What processes can be optimized? How can they be implemented? And what is the impact of set implementations? The goal of this paper is to get insights into the state of the processes and possible optimizations with a LLM.
This paper will analyze the business model and the key sales processes. Then analyzes and filters out the processes, that will benefit the most from optimization with a large language model. After these processes are selected, different models of large language models and tools are introduced and compared. With the local Llama3:8b two prototypes are built with WrenAI and AnythingLLM. These two prototypes are then tested and analyzed with mock data from the client. At the end, the results are discussed and a recommendation is made on how to continue after this paper.
The analysis of the key sales processes has shown that a lot of processes have potential to be optimized with or without the usage of a large language model. A lot of different large language models exist. For this paper the local model Llama3:8b is selected because of security concerns with models that operate in the cloud. For this reason, WrenAI and AnythingLLM, which both operate locally, were tested on the local machine of the author which led to no good results. At this point of time both tools are not recommended to be implemented into the processes of Company XY. With the 70b version of Llama3 the results could already look different. Large language models still need to evolve into better versions to be used to optimize the processes selected in this paper. However, this paper still generated value for the client, who gained insights into its processes and where they can start to work on optimizations even without the usage of a large language model. With this analysis and further improvements in large language models the cornerstone for Company YX is set to be ready when a useable model emerges.
Studiengang: Wirtschaftsinformatik (Bachelor)
Vertraulichkeit: vertraulich
Art der Arbeit
Bachelor Thesis
Auftraggebende
Company XY, Zürich
Autorinnen und Autoren
Kevin Bernet
Betreuende Dozierende
Jüngling, Stephan
Publikationsjahr
2024
Sprache der Arbeit
Englisch
Vertraulichkeit
vertraulich
Studiengang
Wirtschaftsinformatik (Bachelor)
Standort Studiengang
Olten
Keywords
Process optimization, large language model