Leveraging Microsoft 365 Copilot to write Client Project References of Capgemini

Abstract

Generative artificial intelligence (AI) tools, such as ChatGPT and Microsoft Copilot, which are based on large language models (LLMs), are transforming modern work environments by creating new opportunities to enhance productivity. These technologies support tasks such as writing effective text, analyzing documents, and retrieving targeted information from the web. Consequently, organizations are actively exploring use cases to integrate generative AI into their business processes. Nevertheless, they encounter challenges related to cost, scalability, and data privacy during implementation. This research project investigates potential use cases for generative AI in internal processes of the well-established IT consultancy Capgemini. The most promising and feasible use case is implemented while adhering to the company’s generative AI policies. The selected use case addresses writing client project references. These are important to document project experience and outcomes across the organization, serving as valuable resources for future engagements. However, writing these references requires significant time and cognitive effort from employees, resources that are invaluable these days. To streamline this process, Microsoft 365 Copilot is prompted to conduct a structured interview with the employee about the client project. Thereby, Copilot gathers essential information, including the client’s challenges, Capgemini’s solutions, and the resulting benefits. Based on the collected input, Copilot generates the project reference in a predefined format. The prompts used in this process are developed through prompt engineering, an iterative method ensuring that the outputs of the underlying LLM are consistent, relevant, and contextually appropriate. To evaluate the effectiveness and efficiency of the proposed workflow an user study with Capgemini employees is conducted. It examines the perceived effectiveness, efficiency, and user experience of the workflow through hands-on testing and structured questionnaires. The evaluation results demonstrate that the approach is effective and enhances efficiency and user experience in writing standardized, high-quality client project references. However, several improvements possibilities which are valuable for future similar generative AI systems were also revealed. Together, these findings contribute to advancing research on the potential of generative AI to support internal processes in professional environments.

Date of Award	2025
Original language	English
Supervisor	Andreas Stöckl (Supervisor)