Enhancing Citizen Engagement: The Development of Boti, Buenos Aires’ AI Assistant
Introduction
This post is co-written with Julieta Rappan, Macarena Blasi, and María Candela Blanco from the Government of the City of Buenos Aires.
Overview of Citizen Services and AI Integration
The Government of the City of Buenos Aires continuously works to improve citizen services.
Understanding Boti: The AI Assistant
In February 2019, it introduced an AI assistant named Boti available through WhatsApp, the most widely used messaging service in Argentina.
Addressing Bureaucratic Challenges with AI
As Boti grows in popularity, the Government of the City of Buenos Aires seeks to provide new conversational experiences that harness the latest developments in generative AI.
Collaborative Development with AWS
To pilot this new solution, the Government of the City of Buenos Aires partnered with the AWS Generative AI Innovation Center (GenAIIC).
Key Components of the AI Assistant
The solution includes two main components: an input guardrail system and a government procedures agent.
Solution Overview
The Government of the City of Buenos Aires and the GenAIIC built an agentic AI assistant using Amazon Bedrock and LangGraph.
Ensuring Safe Interaction: The Input Guardrail System
Input guardrails help prevent the LLM system from processing harmful content.
Functionality of the Government Procedures Agent
The government procedures agent is responsible for answering user questions.
Innovative Retrieval Methods: The Reasoning Retriever
The agent can use a retrieval tool to provide accurate and up-to-date information about government procedures.
Generating Responses: Answer Generation
After collecting the necessary information, the agent responds using Boti’s distinctive communication style.
Conclusion: The Future of Citizen-Government Interaction
This post described the agentic AI assistant built by the Government of the City of Buenos Aires and the GenAIIC.
Additional Resources
To learn more about Boti and AWS’s generative AI capabilities, check out the following resources.
About the Authors
A section detailing the contributors who collaborated on this project.
Enhancing Citizen Engagement Through AI: The Boti Initiative in Buenos Aires
This post is co-written with Julieta Rappan, Macarena Blasi, and María Candela Blanco from the Government of the City of Buenos Aires.
In a world where technology increasingly shapes our daily interactions, the Government of the City of Buenos Aires is committed to leveraging digital advancements to enhance citizen services. One of their most pioneering efforts is the introduction of Boti, an AI assistant that has transformed the way citizens engage with their local government.
A Revolutionary AI Assistant
Launched in February 2019, Boti is accessible via WhatsApp, Argentina’s most popular messaging service. With Boti, citizens can easily obtain a wealth of information—ranging from renewing driver’s licenses to accessing healthcare services and discovering cultural events. This virtual assistant has quickly become a preferred communication channel, facilitating over 3 million conversations each month.
However, the initial capabilities of Boti were just the beginning. As its popularity surged, the Government recognized the opportunity to enhance user experiences further, particularly in navigating the city’s intricate bureaucracy, which features over 1,300 government procedures, each with its unique nuances.
Addressing Bureaucratic Challenges
The complexity of government processes often poses significant challenges for citizens. Understanding the logic and variations within these procedures can be overwhelming. To tackle this issue, the City Government decided to further empower Boti by integrating generative AI technologies. The partnership with the AWS Generative AI Innovation Center (GenAIIC) aimed at developing an advanced conversational assistant capable of understanding user inquiries more effectively.
This innovative solution includes two critical components: an input guardrail system and a government procedures agent. The guardrail system ensures safe interactions by filtering harmful content, while the procedures agent retrieves pertinent information and formulates responses.
Enhancing Accuracy and Safety
The input guardrail employs a custom LLM classifier to evaluate user inquiries and determine their appropriateness. This is vital, as it enables Boti to provide information relevant to government procedures while blocking harmful requests—like those containing offensive language or unethical prompts.
The government procedures agent leverages a reasoning retrieval system, thus improving the accuracy of information retrieval. This system first identifies and summarizes relevant procedures, then employs a large language model (LLM) to select the best matches for citizen inquiries. This approach has led to significant improvements, achieving as high as 98.9% top-1 retrieval accuracy—a testament to the power of combining advanced AI techniques with practical governmental applications.
Personalized Responses through Lens of Rioplatense Spanish
One of the standout features of Boti is its ability to respond using Rioplatense Spanish, incorporating regional dialects and idiomatic expressions. The AI assistant communicates in a warm and friendly tone, often making use of emojis to enhance interactions. The responses are designed to be concise, direct, and relatable—attributes that resonate well with users.
Subject matter experts involved in the assessment of Boti confirmed the efficacy of its linguistic strategies, recording a 98% accuracy in voseo usage and 92% in using periphrastic futures. These findings emphasize the tailored approach the Government is taking to ensure that digital services are not just functional but also culturally relevant.
Evaluating Performance and Future Enhancements
The strategic partnership between the Buenos Aires government and GenAIIC has produced promising results in making government procedures more navigable for citizens.
Initial evaluations indicated that the guardrails successfully blocked 100% of harmful queries, while the reasoning retriever significantly enhanced retrieval accuracy compared to standard methods by 12.5% to 17.5%. These positive outcomes illustrate the potential for scaling this solution to broader applications within the city.
Looking ahead, the Government of the City of Buenos Aires aims for continuous improvements to Boti. With advancements in generative AI, the introduction of multi-agent frameworks could further enhance interaction capabilities, expanding the services offered and making it even more user-friendly.
Conclusion
The collaboration between the Government of the City of Buenos Aires and GenAIIC is paving the way for a new era in citizen-government interaction. The Boti initiative stands as a distinguished example of how AI can facilitate better communication, foster engagement, and ultimately create a more responsive government.
As the digital landscape continues to evolve, Boti will adapt, ensuring that every citizen has access to the services and information they need—when and how they need it.
To learn more about Boti and AWS’s generative AI capabilities, check out the following resources:
About the Authors
Julieta Rappan, Macarena Blasi, and María Candela Blanco are key figures in the Buenos Aires Government’s digital efforts, driving innovation and enhancing public service delivery. With their extensive experience in digital channels and AI technology, their leadership is critical in transforming the citizens’ experience in Buenos Aires.