Mr. Speaker, please note that the scope of the CRA's artificial intelligence, AI, definition is consistent with the definition of "automated decision systems" outlined in the Treasury Board's directive on automated decision-making. However, the CRA's directive is broader in scope as it includes AI solutions developed for CRA's compliance programs and internal operations in addition to those developed for external service delivery. This definition also includes robotic process automation, RPA, processes that are highly administrative, require little judgment and have clear business rules.
It is important to highlight that the CRA continues to keep humans in the loop of all its AI activities. Human oversight and final decision-making continue to be applied in all types of AI results and program activities.
The CRA does use artificial intelligence in various ways.
The CRA is using AI-based solutions to solve compliance and collection business activities including analysis of patterns, cluster analysis, prescriptive, predictive models and applied predictive analytics i.e., for non-compliance identification, fraud detection, workload selection and compliance strategies.
The CRA employs AI to transform business activities using robotic process automation to automate pre-assessment activities and AI techniques to model and identify processes efficiency gains.
The CRA is using AI-based solutions to transform client service offerings and enhancements through continuous improvements such as the chatbot and improved accessibility. Service improvements are also informed through AI text analytics such as topic modelling, text summarization and sentiment analysis on high volumes of unstructured textual data such as client feedback.
The CRA also uses AI techniques to strengthen data-driven outcomes. Specifically, it is used for research including forecasting, identity and relationship resolution, lead generation and advanced visualization pattern detection. In the research space, the CRA is beginning to experiment with artificial neural networks and recurrent neural networks to test predictive capabilities and assess potential business benefits.
Internally, the CRA uses AI to transform its internal services including natural language processing for analysis of employee surveys. The CRA also uses AI to support security including the evaluation of software/documents to assess their maliciousness, anomaly detection, log collection, tracing and monitoring of accesses.