The OECD.AI Policy Navigator

Our policy navigator is a living repository from more than 80 jurisdictions and organisations. Use the filters to browse initiatives and find what you are looking for.

Modernising CRA’s Service Feedback Program with the integration of generative AI into its national operations


Added by:   OECD analyst
Added on:   29 Oct 2026
Updated by:   OECD analyst
Updated on:   29 Oct 2026

The Canada Revenue Agency (CRA) integrated a generative AI tool into its Service Feedback Program to help employees review cases and prepare correspondence more efficiently. The tool uses internal guidance documents to support daily tasks. It was introduced to reduce manual work, improve quality, and speed up service to clients while keeping employees in control of decisions.

Initiative overview

The Canada Revenue Agency (CRA) has deployed an internal generative AI model as a tool to support employees in the completion of their duties. The Service Feedback Program, with the help of internal service experts in innovation and technology, customised the CRA model for specific implementation in the Agency’s client care program. Prior to business process enhancements and AI deployment, agents were required to manually search within existing procedure manuals for guidance, and to perform manual tasks such as letter creation, case analysis and synopsis. To increase efficiency, customisations related directly to agent responsibilities were added and included adding procedure manuals into the source database and creating dedicated prompts for common Service Feedback Officer tasks.

In an environment characterised by complex procedures, high cognitive load, and diverse client and employee needs, the team standardised and simplified its business processes, successfully transforming an emerging technology into a powerful lever for greater productivity, operational efficiency, and informed decision-making support, with more than 82% of officers using the tool on a weekly basis. Prior to AI deployment, comprehensive testing was undertaken to effectively support employees in its use, and officers are encouraged to use the AI tool, but it remains optional. Feedback from employees has been very positive, with a majority of employees from a recent poll indicating user satisfaction. Initial reviews show a 42% increase in quality, which the program reports is correlated with increased client satisfaction and trust in the organisation.

Before the implementation of AI and other business process re-engineering initiatives, officers typically spent 5 hours per file to complete their tasks. This time has now decreased to 3.75 hours per file, a 25% improvement in efficiency, enabling the program to expedite the service to 10,000 more clients with the same workforce. It is also worth highlighting that, in the preceding year, the program conducted a comprehensive business process review to establish standardised conditions for integrating new technologies. This effort further enhanced efficiency, contributing to an overall efficiency gain of 46% over the past two years.

This initiative demonstrates in concrete terms how technological innovation can be integrated in a sustainable, responsible, and human-centred way within the Agency. The Service Feedback Program has demonstrated through business transformation and the enhancement and integration of digital tools that it is possible to transform a digital tool into a true driver of modernisation, engagement, and support for a national operational program.

Other relevant details

Results, outcomes and impacts: After implementation, average processing time per file decreased from 5 hours to 3.75 hours (a 25% efficiency gain). Quality review scores increased by 42%, which the program reports is correlated with improved client satisfaction. Overall efficiency has improved by 46% over two years, including process redesign. The program can now serve approximately 10,000 additional clients annually with the same workforce. Adoption exceeds 82% of officers weekly and user satisfaction is high. Challenges and lessons learned: Key challenges included integrating AI into complex procedures, ensuring output quality, and maintaining employee trust. The CRA addressed these by first standardising business processes, conducting testing, and making the tool optional rather than mandatory. Clear guidance and training supported adoption. Lessons learned include the importance of process redesign before automation, strong internal governance, and employee engagement from the outset.

About the policy initiative


Category:

  • AI policy initiatives, programmes and projects

Initiative type:

  • AI use cases/projects in the public sector

Status:

  • Active

Start Year:

  • 2025

Target Sectors:


OECD AI Principles:

—

Other relevant urls:

—