North Jakarta City launches ‘local ChatGPT’ to provide public service information

The rise of Generative AI (GenAI) has sparked discussions among Indonesian public sector professionals about the transformative impact it can have and how governments can use it to improve the delivery of public services. 

Speaking at the recent AWS Public Sector Day Indonesia event, former Head of North Jakarta Communication and Information Agency, Yudhistira Nugraha, said GenAI can help the public sector improve the quality of interactions between the government and citizens through personalised digital services.  

One of the initiatives is the launch of jakartautara.info as a platform that aims to help citizens gain access to accurate information more easily. "I call it the ‘local version of ChatGPT’ that will serve the needs of specific information that can only be provided by local government administration," he said.  

The chatbot is designed to answer questions, such as how to make an ID card, apply for birth and death certificates, the location of health facilities, and other public services.   

The platform will also benefit local government leaders, especially those who are new to the organisation, by providing them with the necessary information about their administrative areas, making it easier for them to adapt to their new positions.  

jakartautara.info is a pilot project that can be an example for other local governments in Indonesia who want to adopt GenAI in the government landscape, Yudhistira added. 

To subscribe to the GovInsider bulletin click here.   

Provide information that citizens need 

According to Yudhistira, the idea to create a “local version” of the ChatGPT platform was due to the limited information contained in official government channels. Websites only display general information and do not provide enough details on how to obtain services.  

"jakartautara.info allows citizens to get information about public services in more detail without having to visit government offices," he said.  

The platform can also provide information specific to North Jakarta, which cannot be found on other platforms.   

For example, ChatGPT will not be able to answer questions related to the budget use for a city in detail, or who the authorised officer for certain public services is, because this kind of information will not be available on search engines.   

"By supplying accurate, unique, and contextual data for the model to learn from, this platform will be able to answer questions that ChatGPT cannot," Yudhistira continued.  

The platform will also change the city’s approach to providing information, from government-centric to citizen-centric.  Previously, citizens could only access general information through FAQs with answers prepared by the government. With the use of a large language model (LLM), citizens can ask the government anything they want.    

The importance of data and context  

According to Yudhistira, there are a number of important components for this platform to succeed. The most important thing is training the model on accurate data to prevent the model from providing inaccurate, inconsistent, or misleading information. 

"The assumption is that if the app creator supplies accurate, standardised and contextual data, the model will produce the correct answers," he said.   

To supply such data, organisations can use APIs enriched with additional local data as context.  Data can be inputted in various types of formats, including plain text, datasets in excel, or PDF. This step is also accompanied by continuous testing and validation of information. 

Next is the LLM (large language models). Organisations can finetune existing LLM models such as those provided by AWS or OpenAI or build their own LLM model that suits our needs, he continued.   

Yudhistira emphasised the importance of the commitment of regional leaders to ensure that officials and government personnel provide correct, up-to-date data and have consistent standards.   

In the future, the platform may be upgraded to become the main source of information about the city, beyond information about public services and information about organisational units and personnel. This can be realised if each local government work unit supplies data continuously, he said. 

This article was originally published in Bahasa Indonesia