Indonesia's public sector leaders on what makes AI work in government

Artificial intelligence (AI) gives the Indonesian government an opportunity to go beyond digitising manual processes and rethink how services work. 

In practice, this means helping civil servants process information, identify patterns in data, support policy analysis, and deliver faster, more responsive services to citizens. 

But realising the benefits of the latter requires the government to identify the problems it wants to solve, ensure that the necessary data and infrastructure are in place, and measure whether the technology is making a difference. 

GovInsider speaks to three Indonesian public sector leaders about how AI is being positioned within the digital government, how current data foundations support its use, and how AI’s benefits to citizens can be measured.  

AI must improve the services citizens receive  

Agi Agung Galuh Purwa, assistant deputy for digital government transformation strategy and policy formulation at the Ministry of Administrative and Bureaucratic Reform (PANRB), says the use of AI in government is aimed at improving both the efficiency and quality of public services.  

However, he stresses that a service does not automatically become better simply because it uses AI. The real measure is the impact citizens experience directly.  

“For citizens, what matters more is whether services become easier to access, faster, more accurate, and better able to meet their needs,” he says.  

The boundaries between ministries, agencies, and local governments are not always important from the citizen’s perspective when accessing a service, Purwa adds.  

Citizens want their needs to be resolved without having to understand how government systems work behind the scenes.  

He also explains that the use of AI in public services is part of the shift from the Electronic-Based Government System (SPBE) towards the digital government era.  

Under the Digital Government Master Plan 2025-2045, AI is one of the technologies supporting this transition. Its applications can range from processing information to helping civil servants carry out their work and deliver services.

To achieve the best results, the foundations behind these capabilities first need to be strengthened, including service integration, data exchange, and digital identity.  

AI can then be deployed on top of these foundations to help the government deliver more responsive services, he adds. 

At the same time, Purwa sees tech sovereignty as part of the government’s AI agenda, rather than simply a complementary issue. 

“The main concern is really how this technology can also be developed independently, while ensuring that we have full control over it,” he notes.  

These two objectives need to move forward together, he says. The government also needs to ensure that AI adoption has a clear purpose and does not stop at adding technological features or become simply “AI-washing” in public services. 

Data is the foundation for AI-powered services  

From a technology foundation perspective, Andrari Grahitandaru, head of the Digital Government Research Group at the National Research and Innovation Agency (BRIN), sees AI as having the potential to take government digitalisation to a more advanced stage.

“The biggest opportunity with AI is to move government digitalisation from simply automating processes towards a government that can analyse, predict and provide data-driven recommendations,” she says.  

This opportunity can be applied to a range of government needs, from assisting civil servant, making public information services more accessible, to policy analysis, anomaly or fraud detection, and social interventions.  

Social assistance, healthcare services, and complaints management can also become areas for AI use.  

However, these capabilities do not operate in isolation. AI systems require data that is readily available, shareable, and governed by clear standards. 

According to Grahitandaru, fragmented government data, differences in definitions, data quality issues, metadata, interoperability and access rights can all affect AI’s ability to produce reliable outputs.  

AI therefore needs to be built on a strong, high-quality, and interoperable government data foundation. 

“AI should be a capability built on top of SPBE and Satu Data Indonesia, rather than simply another new application,” she notes. 

With such foundations in place, AI can help the government better understand citizens’ needs and provide more targeted support in both service delivery and decision-making.

Measure impact, not just adoption  

Harnum Annisa Prafitia, BRIN’s Digital Government Research Member, adds that the success of AI in government needs to be assessed based on the outcomes it produces.

“If an application is labelled as AI but does not improve quality, efficiency, accuracy, accessibility or service outcomes, we need to question its added value.”  

Indicators such as service duration, error rates, costs, user satisfaction and the quality of decisions can be used to compare conditions before and after the technology is introduced, she says.  

This measurement also needs to take place throughout the AI lifecycle.  

Testing, validation, documentation, monitoring, and evaluation help the government ensure that systems continue to operate as intended and can be improved when problems arise.  

An enabling ecosystem is also necessary. AI development requires regulation, infrastructure, skilled people, funding, security, and access to computing resources and research data.  

Public service access cannot be overlooked either. AI should expand the government’s ability to serve citizens, rather than become the only gateway to accessing services.  

Non-digital channels remain necessary for people who do not yet have access to, or the ability to use, AI-enabled services, Prafitia adds.  

For citizens, the technology working behind a service is not the main concern. What they need is a government that can resolve their needs more effectively.  

That is where the value of AI in public services needs to be proven.