Governing the Future: Gen AI and Data Governance in Indonesia’s Public Sector

Corinium spoke with Septi Rito Tombe, the senior data manager at INA Digital Edu to understand how Indonesia’s public sector has navigated its digital transformation journey.

Over the past few years, Indonesia’s public sector has steadily navigated its digital transformation journey. By breaking down departmental silos and leveraging centralized data, government agencies are continuously exploring ways to modernize public services and improve efficiency (Lukman & Fiulaizi, 2026). Now, as the digital landscape evolves, the public sector is encountering its next technological shift: Generative Artificial Intelligence (Gen AI) and Large Language Models (LLMs) (IAPA, 2025).

I operate at the center of this shift, balancing a dual role as both a Gen AI lead and a data governance analyst. This creates a professional tug-of-war: my AI role prioritizes agility, massive data ingestion, and automation, while my governance role demands strict access controls, metadata accuracy, and rigorous compliance to prevent data leakage.

Initially, these goals appear conflicting. However, Gen AI models are only as effective and secure as their underlying data. Deploying LLMs alone is insufficient; the public sector must bridge the gap between innovation and data governance to be truly Gen AI-ready. Scaling without a robust framework creates significant operational and security risks. Future success depends on integrating generative power with structural accountability, standardized metadata, and secure data sharing.

Expanding Gen AI-Driven Automation

Gen AI incubation aims to transform static services into proactive, conversational systems, such as LLM-powered assistants for educators. Yet, scaling these requires high data maturity. If underlying data is inconsistent or lacks ownership, AI will hallucinate or expose sensitive information. Innovation must therefore be grounded in a decentralized Data Governance Office (DGO) model that integrates accountability directly into existing organizational functions.

Figure 1: The foundational data pillars required before scaling Gen AI innovation, aligned with industry-standard readiness frameworks.

Data Compliance: The Data Governance Office

Our governance model distributes accountability across three pillars: Data Owners (senior leaders accountable for domains), Data Stewards (experts managing quality and metadata), and Technical Custodians (engineers managing secure access and producing the data). This ensures every dataset feeding a Gen AI pipeline is vetted. Furthermore, we mandate comparative assessments between off-the-shelf models to ensure >90% consistency and cost-efficiency before production.

Figure 2: The Data Governance Office Structure, illustrating the collaborative workflow between business leadership and technical team.

These roles ensure all datasets for Gen AI are accurate and vetted. Accountability also covers algorithms; we mandate comparative assessments of at least two off-the-shelf models per use case. Models for student assessments or teacher chatbots are evaluated against strict baselines: >90% consistency, contextual accuracy, and hallucination rates. We also scrutinize token costs to ensure financial sustainability and budget accountability.

Figure 3: The Gen AI Model assessment, demonstrating how candidate LLMs are rigorously tested for accuracy, consistency, and cost-efficiency before deployment.

Metadata: The Universal Language for Gen AI Interoperability

Gen AI cannot thrive in isolation; it requires cross-sector data interoperability to solve complex multi-domain challenges. When multiple agencies share data to feed a centralized Gen AI model or Retrieval-Augmented Generation (RAG) architecture, semantic misalignment becomes a critical roadblock. If different ministries have varying definitions for a single entity (e.g., what constitutes a "school" or a "civil servant"), the LLM will fail to synthesize the information accurately, leading to severe hallucinations.
This is where centralized metadata management becomes the DNA of Gen AI readiness. Utilizing platforms like OpenMetadata, we create a single source of truth. By implementing a standardized Business Glossary, tracking data lineage, and establishing automated data quality tests, ministries ensure that data definitions are uniform across the board. When a Gen AI model retrieves a dataset to generate an answer, the accompanying metadata provides the necessary context by explaining where the data originated and how to interpret it. This semantic clarity is what transforms raw government data into a powerful, interoperable asset for Gen AI innovation.

Innovation Standards: Balancing Innovation with Compliance

Increased data liquidity for LLMs heightens privacy risks. Gen AI readiness requires strict compliance with UU PDP and the Electronic-Based Government System (SPBE) mandate. Safely grounding models necessitates rigorous classification of data by sensitivity: Public, Restricted, Confidential, or Strictly Confidential, to determine which explorations are fit for production.

Figure 4: Secure Data Classification Workflow demonstrating how we conduct data classification assessments to balance data liquidity for Gen AI with strict privacy compliance.

To practically operationalize these policies and eliminate the risks of ad-hoc AI deployments, we are implementing a Reusable Intelligence Tools platform as our AI gateway. This architecture perfectly resolves our dual mandate: product teams gain the freedom to innovate rapidly, while the centralized AI gateway strictly enforces Role-Based Access Control (RBAC), prompt guardrails, and automated Personally Identifiable Information (PII) redaction before any data ever interacts with an LLM.

Governance as the Catalyst for Gen AI Innovation

There is a common misconception that data governance, with all its rules, protocols, and limitations, slows down the freedom of innovation. In reality, in the era of Generative AI, governance is the very foundation that allows innovation to scale safely. By balancing the drive for Gen AI exploration with the necessary constraints of data governance, we ensure that the conversational agents and generative systems we build tomorrow are not just intelligent, but trustworthy and hallucination-free. The journey toward a Gen AI-driven public sector in Indonesia is a marathon to build a resilient, trusted data infrastructure. Today, strengthening our data governance frameworks, ensuring transparency, security, and compliance, is the ultimate catalyst for safe Gen AI-readiness and transformative public service.

References

• IAPA (Indonesian Association for Public Administration). (2025). Artificial Intelligence in Public Governance: Ethical Opportunities and Challenges in Indonesia's Digital Transformation. IAPA 2025 Hybrid Annual Conference.
• Lukman, J. P., & Fiulaizi, A. (2026). Kebijakan Satu Data Indonesia: Tantangan Dan Peluang Pengambilan Keputusan Publik. Jurnal PubBis, 10(1).

If you are interested to speak at our CDAO and CDAIO series in ASEAN, reach out to Eleen Meleng for more information.