At the recent DBS Asia Future Forward Conference, DBS Group Holdings CEO Tan Su Shan underscored the necessity for businesses to reevaluate their workflows alongside the adoption of artificial intelligence (AI) to ensure meaningful outcomes. Echoing this view, TikTok CEO Chew Shou Zi emphasized the critical role of human judgment in the effective use of AI technologies.

Both executives highlighted an important challenge in AI integration: while companies are adopting AI tools to enhance productivity and reduce costs, there is a growing tension between the expectations placed on employees and their ability to exercise genuine oversight over AI-driven decisions. For instance, an employee tasked with reviewing AI-generated customer application assessments may be pressured to balance careful examination with meeting metrics related to speed and volume of cases processed.

This dynamic raises concerns about whether employees truly have meaningful control over AI-assisted decisions. Careful assessment and questioning of AI recommendations often require additional time and deliberation, which may conflict with organizational targets that prioritize efficiency. Consequently, employees might feel compelled to approve AI outputs to meet performance goals, even when doubts arise, undermining the intended safeguard of human oversight.

Experts note that this situation reveals a gap between the principle of human oversight and its practical implementation. When errors occur in AI-assisted decisions, the immediate attribution of responsibility to individual employees overlooks whether they had adequate time, access to information, relevant expertise, and the authority to challenge or escalate problematic recommendations.

To address these issues, organizations must establish conditions that enable meaningful employee oversight over AI systems. This includes allocating sufficient time for employees to review AI outputs carefully, providing training to identify potential errors, and fostering an environment where staff can reject or escalate questionable AI decisions without fear of penalties. Furthermore, accountability should extend beyond individual workers to encompass the organizations that select and deploy AI systems and set operational targets that influence employee behavior.

As Singapore advances its push for broader AI adoption, businesses are encouraged to go beyond assessing whether AI can make processes faster or cheaper. Instead, emphasis should be placed on whether AI integration enhances decision quality in practice and empowers employees to scrutinize AI recommendations effectively when irregularities arise.

While AI has the potential to increase work speed, experts caution that faster does not necessarily equate to better. For individuals held accountable for decisions assisted by AI, genuine oversight requires more than just the ability to approve recommendations; it necessitates real authority and capacity to question and challenge AI outputs.