The ROI of Responsible AI: Governance as a Competitive Advantage

Written by Zora Salam (W’30) & Edited by Sylvia Yum (C’30)

In 2024, an Air Canada passenger asked the website’s chatbot about a bereavement fare, a specialized flight discount for when an immediate family member passes away. The chatbot told him that he could apply for the discount after booking the full-priced tickets. This information was wrong, and when he requested a refund, Air Canada denied it. He then took it to a tribunal. Air Canada argued that the chatbot was a separate legal entity, responsible for its own actions.

The tribunal found this argument unconvincing and ruled that, because the chatbot was part of the airline’s website, Air Canada was responsible. Even though the damages totaled only around CA$650, the real cost was far greater: the case showed that AI is now allowed to execute decisions without human approval. Deloitte’s latest survey found that 40% of companies let AI execute decisions without human approval. Furthermore, the issue is compounded with nearly one in four saying accountability is unclear when AI produces a bad outcome. The real issue is that companies are giving AI authority without a clear responsibility check.

This is increasingly evident in hiring, creating an “AI arms race” where both job seekers and employers are using AI tools against each other. According to Reuters, more than 80% of U.S. employers use AI in hiring. This is a deeper issue with accountability, as when automated tools are being used on both sides, a decision can pass from one algorithm to the next without a person ever being responsible for it.

Candidates are also beginning to use AI to get past AI screeners, with employers using AI to sort large piles. In Clutch’s survey of job seekers, 80% said they use AI in their search, 86% said it helped them submit more applications each week, and 77% believed their applications were screened using AI. This then leads to the issue of candidates being rejected without a single person ever reading their application. This was exemplified in the Mobley v. Workday case, where an applicant is suing Workday’s AI screening software for discriminating against him because of his age and disability. While these allegations were left unproven and Workday’s claim that lawsuit is without merit, a federal judge in June refused to dismiss most of the discrimination claims. A key consideration is the fact that Mobley sued the hiring software, not the employers who used this software. When a hiring decision involves an employer, a software vendor, and an algorithm, it is unclear who is responsible. 

Companies can settle this question before a court does, and that is the purpose of AI governance. This means assigning ownership of each AI system, deciding what needs human sign-off, and testing and recording outputs. Companies that are taking this most seriously are seemingly doing better. In Protiviti’s 2026 survey, 63% of boards at orgnaizations with high AI returns put AI on the agenda at every meeting, relative to the just 13% of low-return organizations. 

There are limits however. Human review slows things down. This data could also just be correlation, not causation: companies with strong AI returns may simply have more resources for governance. As AI makes bigger decisions about who gets hired or approved, the cost of responsibility grows. The ones that decide responsibility in advance will earn an important advantage: trust. As everyone begins to possess the same tools, the real return on responsible AI is being the company, companies, or candidates can trust.