Initiative overview
The Canada Revenue Agency (CRA) has deployed a publicly accessible generative AI chatbot to support taxpayers in navigating tax-related information and services. The chatbot is designed to answer common, high-volume questions using authoritative CRA web content, helping users find accurate information more quickly while reducing pressure on traditional service channels.
During its initial deployment period, the chatbot handled over one million chat sessions and responded to more than 1.5 million taxpayer questions, representing a significant increase in digital self-service interactions year-on-year. While the CRA does not attribute changes in call volumes directly to chatbot usage, the scale of adoption suggests meaningful demand for AI-enabled digital service delivery.
The solution is designed to be modular and scalable, with strong controls to ensure transparency, appropriate content sourcing, and alignment with public-sector values. Human oversight, content governance, and clear user disclosures are integral to the design. This use case demonstrates how generative AI can enhance public service delivery while maintaining trust, accountability, and accessibility.
Results, outcomes and impacts
During its initial deployment, the chatbot handled over 1 million chat sessions and responded to more than 1.5 million questions. Uptake was measured through usage analytics, session volumes, and interaction counts. While call volume changes cannot be directly attributed to the chatbot, adoption levels indicate strong demand for digital self-service.
Future impacts are expected to include improved access to information, increased service efficiency, and continued growth in digital engagement.
Challenges and lessons learned
Key challenges included ensuring accuracy, managing risks related to generative AI responses, and maintaining public trust. The CRA addressed these through clear user disclosures, limiting the chatbot to official web content, human oversight, and ongoing monitoring. Another challenge was setting appropriate expectations about the tool’s capabilities and limits.
Lessons learned include the importance of strong governance from the outset, transparent communication with users, and phased implementation. Early monitoring and feedback mechanisms are critical to identify gaps and improve performance. Success depends on clear scope definition, senior leadership support, technical safeguards, and alignment with public service values such as fairness, accountability, and accessibility.
