The new comprehensive series on AI's evolution

From CPA Canada

CPA Canada’s new three-part series aims to strike a balance of innovation with ethical responsibility when it comes to Artificial Intelligence in the industry

CPAs find themselves at the juncture of technological disruption and ethical responsibility when it comes to the application of AI within the accounting profession. In collaboration with the American Institute of CPAs (AICPA), CPA Canada has issued its first publication in a new three-part series addressing the profound impact of AI, and offering trusted guidance for navigating the intricacies involved, from deciphering machine learning algorithms to ensuring transparency in AI-driven decision-making.  

“What I hope comes from this paper is an enablement of cautious experimentation,” says Melissa Robertson, principal of Research & Thought Leadership (Technology) at CPA Canada. “When fewer people are afraid of AI, the more willing they will be to utilize new tools and find meaningful opportunities to use them.”  

It’s crucial for CPAs to grasp the intricacies of AI technology to ensure its effective and ethical use within accounting, safeguarding against potential errors and biases while maximizing its transformative potential. To brief readers, the paper begins by addressing the technology’s most promising advancements and where they currently fall short.  

Generative AI 

Over the past couple of years, the landscape of AI has been dramatically reshaped by a series of innovations. One such innovation is generative AI—a recent form of technology that has stretched beyond the capabilities of conventional AI, moving from analyzing data to creating entirely new content. 

Some of the use cases of generative AI for accountants include financial report generation, forecasting, predictive analysis and compliance. However, there are challenges such as ensuring the accuracy and reliability of AI outputs and addressing biases in AI systems.  

“Regardless of whether you’re using AI or any automation technology, there is absolutely still the same requirement to be able to demonstrate completeness and accuracy,” says Robertson. “In traditional technologies, such as simple algorithms or calculations, verifying accuracy and completeness is relatively straightforward by reviewing source code or recalculating. However, auditing AI algorithms is difficult due to their complexity, and there is currently no standardized method for doing so.” This can impact the ability to confirm the accuracy and completeness of AI-generated outcomes, highlighting the need for new approaches to auditing AI tools. She adds, “In many ways, I consider the need for AI system auditing to open new career paths for CPAs and the technology experts they will be working closely with.”  

Foundation Models and General-Purpose AI Systems (GPAIS) 

Foundation models like GPAIS can perform various tasks across different domains. Foundation models are like the foundation of a house, providing a solid base upon which developers can build diverse AI applications. Yet, while GPAIS offers versatility, it may lack the specificity required for certain tasks, impacting precision and detail in outputs.  

Multimodal Generative AI 

Generative AI demonstrates its full capabilities when it can interact with data in diverse formats. For instance, if an accountant employs generative AI to review various types of information and audit evidence during an audit, the AI could analyze financial text records, numerical data, visual charts and even evaluate video testimonials. 

While multimodal AI systems are still in their infancy, it’s anticipated over the next year to proliferate and expand the capabilities of GPAIS. However, it poses challenges such as increased complexity and potential errors which accountants must address through governance and control programs.  

Regulatory Challenges and Responses 

Regulators and standard setters struggle to keep up with the advancements in AI, leading to challenges in regulating its development and use. “There is not a ton of regulation right now, and I think we're in a very early stage here in Canada,” says Robertson, emphasizing a growing need for advocacy and community awareness. 

What comes next? The second paper in the series between CPA Canada and AICPA is poised to further examine the potential risks and impacts of AI, and leading practice governance and control practices. It will also examine the role of guidelines, standards, and regulations in establishing and managing a robust responsible AI program. 

Read the full white paper and learn more about CPA Canada’s series on Artificial Intelligence here.