Artificial intelligence experts have raised concerns regarding the Canada Revenue Agency (CRA) and its plan to utilize AI for assisting Canadians with tax-related inquiries. They argue that it is premature for the CRA to adopt AI solutions, given that human agents at call centers have frequently failed to deliver accurate information.
In a recent report, federal auditor general Karen Hogan highlighted significant shortcomings in the CRA”s call response capabilities. The report indicated that the agency answered a minimal number of incoming calls, and when it did respond, less than one in five callers received correct information regarding personal income taxes. Hogan”s findings revealed that “Charlie,” a rule-based AI chatbot, provided accurate responses only about a third of the time. Unlike generative AI that creates new content based on queries, Charlie pulls information from predefined scripts designed to address common questions.
The CRA is currently testing a generative AI version and is extending the operational hours of its chatbots, as well as increasing the range of questions they can handle. However, experts emphasize that the CRA must first rectify issues with human agents before fully implementing AI technologies.
Anatoliy Gruzd, a Canada Research Chair in Privacy Preserving Digital Technologies, cautioned that reliance on AI should be reconsidered until human performance is improved. He stated, “If you”re a government agent… you want to make sure that you have the process nailed down with the human agent.” Gruzd further explained that the effectiveness of chatbots is heavily reliant on the accurate data provided by human agents, noting that understanding the sources of human error is essential before launching extensive AI initiatives.
Adegboyega Ojo, a professor at Carleton University specializing in AI governance, suggested a balanced approach integrating both human and AI capabilities. He argued that while machines can handle straightforward inquiries, more complex and nuanced questions should be directed to specialists. Ojo emphasized, “When you have complex tasks… this is why you need a machine,” highlighting that AI can alleviate pressure from high call volumes but should not replace human workers entirely.
Furthermore, Jasmin Manseau, an associate professor at the University of Ottawa“s Telfer School of Management, noted that AI excels in predictive tasks rather than judgment-based ones, acknowledging the potential for machine errors. He stated that the government”s exploration of AI is vital for advancing technology, yet public trust is crucial for successful implementation. “If it is accurate and people trust it, then they”ll use it,” he remarked, stressing the importance of reaching high accuracy thresholds for public acceptance.
This report, originally published by The Canadian Press, underscores the need for the CRA to address its human response issues before fully embracing AI technologies in its operations.
