Thirty seven percent of Singapore organisations are scaling AI across multiple functions, but only 20 percent say they are highly prepared to manage autonomous AI deployment — the gap that Insight calls the 'autonomy paradox'.
Thirty-seven percent of Singapore organisations are scaling AI across multiple functions, but only 20 percent say they are highly prepared to manage autonomous AI deployment, according to an Insight survey of 538 business decision-makers in Singapore and Australia (ITBrief). The gap is what Insight, the consultancy running the survey and selling AI advisory work alongside it, calls the 'autonomy paradox': the more decision-making latitude companies hand to AI, the more they need oversight structures that are still being drawn up.
Fifty-five percent of Singapore leaders report moderate-to-strong returns on their AI investments, against 41 percent of Australian peers in the same study. Fourteen percent of Singapore organisations say they have fully embedded AI into operations. Both numbers frame why Singapore is moving ahead, and why the paradox compounds there first: the ROI case is being made on the very systems whose behaviour is least scrutinised, which pulls deployment further ahead of the controls meant to constrain it.
Readiness sits on the other side of the same ledger. Twenty percent of Singapore respondents describe themselves as highly prepared for autonomous AI deployment, and nearly 40 percent describe themselves as only moderately prepared — two distinct groups that together make up the larger half of the sample. The 14 percent who report full AI integration are, by their own account, the most exposed. Both figures are self-reported perceptions from one consultancy's sample, not audits, and the same caveat applies to the ROI numbers above. Read together they sketch a workforce that says it is winning on AI output while flagging it has not finished building the supervision.
Singapore's regulatory infrastructure is moving, if not at the same pace. The Info-communications Media Development Authority (IMDA) released the second edition of its Model AI Governance Framework in 2024, the most concrete national anchor for the guardrail side (IMDA). The framework sets out guidance on internal controls and human oversight of AI systems, but uptake is voluntary and the survey suggests many firms have not yet wired it into operating practice. The asymmetry Insight flags, in other words, lives on both sides: firms racing to deploy, and a regulator offering the playbook those firms have not yet started following.
The practical read is not that firms should slow deployment; the ROI numbers already punish that posture. Governance has to be specified and budgeted alongside the agents and models it monitors, with audit, escalation and rollback defined up front rather than bolted on after an AI workflow is already running in production. Insight's named fix is to build oversight in from day one, not retrofit it. The deployment data is what turns that advice from a slogan into a deadline.
The next milestones worth watching are whether IMDA moves its voluntary framework toward audited or mandatory expectations in its next work programme, and whether the next Insight survey shows the readiness gap narrowing or widening after another year of AI deployment at scale.