Survey Finds Agentic AI Audit Gap in Singapore Firms

Singapore businesses are racing to deploy agentic AI — systems that can plan and execute multi-step tasks without human hand-holding — but the governance scaffolding isn't keeping pace. A new survey by identity verification firm Sumsub and the Singapore FinTech Association found that while 94% of respondents are using or piloting agentic AI, barely more than a quarter can explain what those systems actually did. The numbers paint a picture of an ecosystem that's sprinting forward on deployment while walking on accountability.

The survey, conducted by Blackbox Research between April and June 2026 across 720 AI decision-makers in nine APAC markets, found that only 29% of Singapore respondents could produce an audit trail of their AI systems' decisions. About 70% have guidelines assigning direct responsibility for AI outcomes — 40% to a specific person, 30% to a team — but responsibility without traceability is a paper shield. Singapore scored 65.6 on the study's AI governance benchmark, slightly below the regional average of 67.1. Sumsub attributed the gap partly to Singapore's more advanced regulatory framework, which may lead firms to rate themselves against stricter standards.

The hesitancy shows up in other numbers too. Only 16% of Singapore respondents had significantly expanded the scope or autonomy of their AI systems in the past year — the lowest rate among APAC markets surveyed. Businesses were far more comfortable letting AI handle routine tasks (90% willing) than anything involving financial liability. The main barriers cited were the technical complexity of modern AI models (66%), platform integration challenges (50%), and difficulty tracking actions taken by external AI tools (49%).

SFA president Holly Fang said businesses need to strengthen traceability, accountability, and governance as AI moves from assisting workers to handling critical workflows autonomously. The study found that 98% of Singapore respondents said they were ready to adopt third-party verification systems that link autonomous AI actions to verified identities — a striking near-consensus that suggests the market knows the gap exists and is looking for tools to close it. Data-related tasks saw the biggest AI impact at 29%, followed by operations and workflow processing at 21%, with fraud detection and anti-money laundering at 15%.

Why it matters for Singapore: When 94% of firms are using agentic AI but only 29% can audit it, the gap isn't theoretical — it's a compliance liability that regulators are watching. Singapore's MDDI has already introduced a Model AI Governance Framework for Agentic AI, and the Monetary Authority of Singapore has been steadily tightening expectations around model risk management. The survey suggests the market is eager for verification tools, but tooling alone won't solve the problem if responsibility assignments remain split between individuals and teams without clear escalation paths. The firms that close this gap first will have a regulatory advantage when the frameworks become requirements.