Optimizing Skill Management at Siemens with Generative AI for Enhanced Workforce Development

Dijin David Thomas

Diplomová práce

Anotace:

This thesis examines how generative AI can close the competence-visibility gap in corporate learning ecosystems, where evaluation by self-report and static document retrieval fails to capture authentic employee capability. Siemens MyLearningWorld provides the case-study context. The work asks how an agentic conversational layer added to such a platform can make competency assessment continuous, transparent, and accountable to the learner. The study designs, implements, and evaluates a multi-agent prototype in which specialised agents handle assessment, mentoring, recommendation, and progress tracking through a single chat interface. The agents apply a three-dimension competency rubric, persist results as a longitudinal learner record, and operate within granular consent and explicit data-isolation controls. The prototype was built through an AI-assisted development method that takes domain expertise as the primary input rather than manual coding. A mixed cohort of evaluators drawn from external and host-organisation respondents rated the prototype's perceived usability in the highest band of a standard usability instrument. Independent stakeholder engagements in private-sector and public-sector contexts each pointed to concrete pathways for integration. The work contributes an evaluated agentic-upgrade pattern for enterprise learning platforms, integrating conversational competency assessment, longitudinal record-keeping, and regulation-aware governance. Findings indicate that the integration pathway into the host platform is viable subject to organisational review.

Abstract:

This thesis examines how generative AI can close the competence-visibility gap in corporate learning ecosystems, where evaluation by self-report and static document retrieval fails to capture authentic employee capability. Siemens MyLearningWorld provides the case-study context. The work asks how an agentic conversational layer added to such a platform can make competency assessment continuous, transparent, and accountable to the learner. The study designs, implements, and evaluates a multi-agent prototype in which specialised agents handle assessment, mentoring, recommendation, and progress tracking through a single chat interface. The agents apply a three-dimension competency rubric, persist results as a longitudinal learner record, and operate within granular consent and explicit data-isolation controls. The prototype was built through an AI-assisted development method that takes domain expertise as the primary input rather than manual coding. A mixed cohort of evaluators drawn from external and host-organisation respondents rated the prototype's perceived usability in the highest band of a standard usability instrument. Independent stakeholder engagements in private-sector and public-sector contexts each pointed to concrete pathways for integration. The work contributes an evaluated agentic-upgrade pattern for enterprise learning platforms, integrating conversational competency assessment, longitudinal record-keeping, and regulation-aware governance. Findings indicate that the integration pathway into the host platform is viable subject to organisational review.

Klíčová slova

Retrieval-augmented generation, Action research, System Usability Scale, Siemens MyLearningWorld, EU AI Act, Agentic AI, Multi-agent systems, Competency assessment

Keywords

Agentic AI, Multi-agent systems, Retrieval-augmented generation, Competency assessment, Action research, EU AI Act, System Usability Scale, Siemens MyLearningWorld

Jazyk práce: angličtina

Datum vytvoření / odevzdání či podání práce: 3. 5. 2026

Obhajoba závěrečné práce

• Obhajoba proběhla 3. 6. 2026
• Vedoucí: Pavel Sládek
• Oponent: Miloš Maryška

Vysoká škola ekonomická v Praze

Magisterský studijní program / obor:
Information Systems Management / Management of Business Informatics