AI Agents: Accountability & Oversight in Business

Asia Pacific businesses are now working with two workforces: their people and AI agents. Companies like Singtel are redrawing their organization charts to reflect this shift. An AI agent doesn't just answer questions; it acts on them, like approving an invoice or updating a customer record. This capability makes agents valuable but also risky. They do exactly what they're told, without instinct for when something goes wrong. The challenge comes with accountability. If an agent makes an error, the manager who deployed it is held responsible, even if they cannot see what the agent did. There's also a risk of agents drifting beyond their assigned tasks, potentially accessing unauthorized data or causing compliance problems. Another concern is "AI sprawl," where teams create many agents without central oversight. This is similar to "shadow IT" but with higher stakes because agents can act independently. The bottom line: While agents offer real productivity benefits, leaders must prioritize oversight. The key question is not how many agents a business can deploy, but whether it can constantly monitor what each agent is doing and identify the person accountable for it.