From Prompt to Page : Exploring the Use of Large Language Models to Support Technical Documentation at Toyota Material Handling Manufacturing Sweden
Year: 2025
Type: dissertation
Abstract: This study explores how large language models (LLMs), specifically Microsoft Copilot, can support technical documentation work at Toyota Material Handling Manufacturing Sweden (TMHMS). The study focuses on two selected use cases: text generation and translation quality control. Technical writers evaluated the performance of different prompting techniques on these use cases. The study was conducted with a mixed-methods approach, using both semi-structured interviews and an exploratory prompt development process. The findings indicate that it is possible to create prompts that yield acceptable and useful outputs within tasks such as text generation and translation quality control. Persona prompting emerged as the most preferred technique for text generation, while reflection prompting was most favored for translation quality control. These findings also highlight that users’ trust in the AI system plays a key role in their evaluation of usefulness and satisfaction. Prompt engineering proved useful in general contexts but showed some limitations when applied to highly technical language. The results suggest that Copilot when combined with prompting strategies, represent a promising technique for supporting technical documentation work. For stronger domain alignment with TMHMS’s language, terminology and data, more advanced integration methods such as fine-tuning, domain-adaptive pre-training, or retrieval-augmented generation (RAG) could offer added value. However, further investigation is needed to determine their feasibility and value.
Source: DiVA (Linkoping University)
Author: Törby, Maria
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7052058257