Analysis, planning, optimization, and automation of manual processes within Amazon

This thesis addresses the challenge of rising operational workload in large enterprises and the difficulty of scaling traditional automation beyond pilots. From the theoretical background, it highlights the evolution toward reasoning-capable or agentic AI, which combines autonomy, adaptability, and...

Descripción completa

Detalles Bibliográficos
Autor: Gibert Morera, Jan
Tipo de recurso: tesis de maestría
Fecha de publicación:2025
País:España
Institución:Universitat Politècnica de Catalunya (UPC)
Repositorio:UPCommons. Portal del coneixement obert de la UPC
Idioma:inglés
OAI Identifier:oai:upcommons.upc.edu:2117/444802
Acceso en línea:https://hdl.handle.net/2117/444802
Access Level:acceso embargado
Palabra clave:Artificial intelligence--Financial applications
Production control -- Data processing
Product management -- Automation
Quality of products
Intel·ligència artificial--Aplicacions a les finances
Producció -- Control -- Automatització
Gestió de productes -- Automatització
Qualitat dels productes
Àrees temàtiques de la UPC::Economia i organització d'empreses
Descripción
Sumario:This thesis addresses the challenge of rising operational workload in large enterprises and the difficulty of scaling traditional automation beyond pilots. From the theoretical background, it highlights the evolution toward reasoning-capable or agentic AI, which combines autonomy, adaptability, and governance to move from isolated task automation to end-to-end process integration. The project, conducted during a six-month internship at Amazon, applied these principles in practice: more than one hundred manual processes were mapped, scored by impact, priority, and complexity, and distilled into a portfolio of sixty opportunities. Pilot automations validated that agentic AI could reduce manual effort while maintaining reliability and gaining user acceptance. Building on these pilots, a three-year roadmap was developed projecting up to 80% workload automation and significant operating expenditure savings. Results align with industry benchmarks, confirming that leadership sponsorship, value frameworks, and workforce engagement are essential for success. For confidentiality reasons, process details are anonymised and all outcomes are expressed in percentages. The thesis demonstrates both the potential and the dependencies of scaling reasoning-capable AI in enterprise contexts.