Publicação
Evaluating No-Code AI Knowledge Assistants: Vector Database-Driven Chatbots for Digital Transformation in Enterprises
| datacite.subject.fos | Ciências Naturais::Ciências da Computação e da Informação | |
| datacite.subject.sdg | 08:Trabalho Digno e Crescimento Económico | |
| datacite.subject.sdg | 09:Indústria, Inovação e Infraestruturas | |
| datacite.subject.sdg | 12:Produção e Consumo Sustentáveis | |
| dc.contributor.advisor | Lopes, Nuno Alexandre Moura Pinto | |
| dc.contributor.author | Nienaltowski, Maciej | |
| dc.date.accessioned | 2026-07-03T09:16:04Z | |
| dc.date.available | 2026-07-03T09:16:04Z | |
| dc.date.issued | 2026-06-24 | |
| dc.description | Dissertation presented as the partial requirement for obtaining a Master's degree in Information Management, specialization in Digital Transformation | |
| dc.description.abstract | This thesis examines whether no-code platforms can support the development of effective knowledge assistants for enterprise environments by designing and evaluating a Retrieval-Augmented Generation (RAG) system. The study addresses the problem of information fragmentation, where employees struggle to find reliable guidance across dispersed documents. The prototype integrates workflow orchestration, cloud document monitoring, vector database storage, and language models to automate document ingestion, semantic retrieval while excluding archived content, and response generation with safeguards against unrelated queries. The system was evaluated using 24 UK government policy documents and 21 structured queries covering factual recall, synthesis, negative testing, and document lifecycle scenarios, achieving an average accuracy score of 4.67 out of 5.0 and a relevance score of 5.0 out of 5.0 via LLM-as-judge evaluation. The design includes continuous synchronization between cloud storage and the vector database, metadata-driven document management, a two-stage retrieval process combining vector search with semantic prioritization, and structured error logging, all implemented through visual workflows without custom code. The results indicate that such systems can be implemented using no-code tools under controlled conditions, although limitations remain regarding optical character recognition for scanned documents, role-based access control, and validation in multi-user settings. The study suggests that design choices related to orchestration, metadata structure, and document lifecycle management play an important role in system performance alongside the underlying language models, and it provides a reproducible approach for developing enterprise knowledge systems using visual development tools. | eng |
| dc.identifier.uri | http://hdl.handle.net/10362/204297 | |
| dc.language.iso | eng | |
| dc.rights.uri | http://creativecommons.org/licenses/by/4.0/ | |
| dc.subject | No-Code | |
| dc.subject | Artificial Intelligence | |
| dc.subject | Digital Transformation | |
| dc.subject | Retrieval-Augmented Generation | |
| dc.subject | Vector Database | |
| dc.subject | Knowledge Management | |
| dc.title | Evaluating No-Code AI Knowledge Assistants: Vector Database-Driven Chatbots for Digital Transformation in Enterprises | eng |
| dc.type | master thesis | |
| dspace.entity.type | Publication | |
| thesis.degree.name | Mestrado em Gestão de Informação, especialização em Transformação Digital |