A Contextual AI Agent Decision Support System for e-Government Cybersecurity Self-Assessment
Year: 2026
Type: conference-paper
Abstract: The rapid digitisation in Indonesia's public services poses various risks that force institutions to conduct a regular cybersecurity self-assessment. Nonetheless, traditional self-assessment may be insufficient because it results only in a rigid compliance score, without contextual recommendations based on institutional conditions. Although the advancement of generative artificial intelligence (GAI) offers significant potential to automate decision support systems (DSS), its implementation is often evaluated under generic models with zero-context regarding institutional conditions, resulting in hallucinated idealistic solutions that exceed actual institutional resource capacities. This study proposes a contextual AI agentbased DSS that dynamically injects organisational context at the prompt level to mitigate irrelevant recommendations. This work conducts a scenario-based empirical simulation across six distinct scenarios, using an asymmetric baseline matrix from prominent AI agent architectures, including Gemini, OpenAI and Meta, under both zero-context and context-aware conditions. The results show that context injection can optimise decision support outcomes by guiding model reasoning from generic suggestions toward more relevant, realistic and affordable action plans. Particularly, this framework could bridge the gap between regulatory compliance and operationally realistic conditions by enabling resourceconstrained government institutions to execute a feasible cybersecurity effort.
Institution: National Research and Innovation Agency
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7220998660