Noah Tesfaye on How AI Is Reshaping Audit and Ethical Leadership
Until recently, the practice of audit relied on samples.
Auditors would select a small group of transactions from a larger population, test them carefully and use professional judgment to draw their conclusions. It was the most effective way to manage risks with the tools that were available.
“I remember thinking that we were making judgments about an entire population based on a tiny window into it,” says Noah Tesfaye, a Senior Auditor at the Office of the Auditor General of Canada.
However, today’s auditors are facing a different reality.
“With AI, we now have the opposite challenge,” he says. “We can see everything. Full ledgers, complete data sets, entire populations of transactions. But still, we do not always understand why things look the way they do.”
The shift from limited to overwhelming visibility is fundamentally changing the practice of the audit.
The huge uptick in available data combined with AI tools is helping auditors cut down on time-consuming work, leaving more time for critical analysis.
The same evolution is happening with document review and administrative tasks: things that once ate up hours of an auditor’s day are becoming more streamlined.
But while AI is improving efficiency, Noah cautions that the profession must be careful not to confuse speed with certainty.
“Technology should inform judgment, not replace it.”
“A big misconception is that AI is just another tool like Excel,” he says. “It is not. AI influences decision-making. And because of that, it can scale both efficiency and bias.”
Unlike traditional software, AI systems are often open-ended and adaptive. They generate outputs based on historical training data, which can pose a number of ethical risks.
Noah groups these risks into three key areas: bias at scale, opacity and accountability gaps.
Bias at scale occurs when flawed assumptions within training data are replicated and amplified through AI-generated outputs. Think of what happens when you have a crack in the foundation when building a tower: it creates a weakness that threatens the entire structure.
Opacity refers to the difficulty of understanding how some AI systems arrive at conclusions, particularly when their inner workings function as “black boxes.” Without being able to see how an AI system got from point A to point B, it’s hard to trust that it hasn’t simply hallucinated its output.
Accountability gaps emerge when it becomes unclear who is ultimately responsible for decisions influenced by AI systems. When misinformation is generated by an AI system but used by a human, there tends to be a certain amount of assigning blame.
“When a system heavily influences a decision, who is accountable?” Noah asks. “The developer? The manager? The auditor? AI itself is not a legal entity.”
Accountability ultimately falls on the CPA responsible for the output. CPAs have been advocating for a ‘human in the lead’ approach to AI use, ensuring responsibility and preventing over-reliance.
For Noah, these questions about accountability make ethical leadership more important than ever.
“Ethical leadership in the age of AI is not about having all the right answers,” he says. “It is about designing systems and cultures where questioning is built in... technology should inform judgment, not replace it.”
Noah believes that CPAs have a critical role to play in shaping how these technologies are governed and applied responsibly.
“It is a pivotal moment,” he says. “The ethical principles we already have matter more than ever before. The challenge now is learning how to apply them in an environment that is becoming increasingly automated, data-driven and complex.”
Noah Tesfaye will be presenting at the CPA Ontario Ethics Conference: Human-Centred Leadership in the Age of AI on June 18.