Singapore’s data professionals are proving to be among the most cautious in the world when it comes to letting artificial intelligence (AI) operate unsupervised.
According to a new global study, 61 per cent of the city-state’s data analysts prefer a human-in-the-loop approach to AI, the highest share recorded across all regions surveyed.
Also Read: AI in Singapore: From generative tools to real-world impact
The finding comes from Alteryx’s “2026 State of the Data Analyst: The Rise of Business Logic” report, which polled 1,400 respondents worldwide, including 175 data analysts and IT leaders in Singapore. It paints a picture of an ecosystem that is deploying AI aggressively, yet remains deeply uneasy about handing it the keys.
Just 1 per cent of Singapore respondents said they were comfortable with AI operating fully autonomously, a third of the already-slim 3 per cent global average. For a market that has positioned itself as Southeast Asia’s AI testbed, courting hyperscaler data centres and government-backed AI programmes, the reluctance is telling. It suggests that enthusiasm for AI adoption at the policy and infrastructure level has not necessarily trickled down into trust at the operational level, where analysts are the ones left cleaning up after the models.
Strategic weight is rising, but so is the workload
The report does not describe a market turning its back on AI. Quite the opposite — 75 per cent of Singapore respondents said AI’s strategic impact on their organisation has grown over the past year, as companies lean further into automation and agent-based systems. And 66 per cent agreed that AI and agent-based systems perform best when managed at the business-unit level rather than by centralised data or IT teams, a preference that echoes a broader shift happening across the region, where domain teams increasingly want ownership over the tools shaping their decisions, rather than waiting on a central function to translate their needs.
That shift, however, is generating friction of its own. Singapore’s analysts are spending significant chunks of their working week doing the unglamorous groundwork AI still cannot do reliably alone: an average of five hours a week preparing and cleaning data, and a further three hours correcting and validating AI-generated outputs. Put together, that is roughly a full working day each week spent making sure the machine’s homework is actually right.
Data quality, not the models, is the real bottleneck
Perhaps the most pointed figure in the report is this: 46 per cent of AI and analytics projects in Singapore that fail to meet their objectives are attributed primarily to data-related issues, rather than problems with the underlying models or tooling. In other words, the technology is rarely the weak link; the data feeding it is.
Also Read: Singapore turns AI scrutiny towards chatbots, personal data, and digital twins
This tracks with a pattern seen repeatedly across Southeast Asia’s broader digitalisation push, where legacy systems, fragmented data ownership across departments, and inconsistent data hygiene practices have quietly undermined more ambitious AI rollouts. Singapore, despite its relatively mature digital infrastructure compared with regional peers, is not immune.
The report also flags governance as a growing pain point sitting alongside the data quality problem. Thirty-seven per cent of respondents cited data quality issues as a leading source of friction when deploying AI, while 38 per cent pointed to data access approvals, the bureaucratic back-and-forth of getting the right people cleared to use the right datasets. Meanwhile, 46 per cent said unclear ownership and accountability for AI-driven decisions is a barrier standing between generating an AI insight and actually being able to act on it.
Taken together, the figures describe an organisational bottleneck as much as a technical one. Companies can buy the AI tools, but if nobody is clearly responsible for the decisions those tools inform, and if data access still requires multiple rounds of sign-off, the promised speed gains from automation start to erode.
“Business logic” as the missing layer
Philip Madgwick, Alteryx’s regional vice-president for Asia, framed the findings around a familiar tension: the gap between deploying AI and actually trusting what it produces. He said many organisations in Singapore are still contending with poor data quality, weak governance and lingering uncertainty over AI-generated outputs, and that what separates companies that pull ahead from those that stall is whether the people closest to the business are the ones defining and managing the logic behind AI’s decisions.
That framing matters for how Southeast Asian companies think about their next phase of AI investment. Much of the region’s AI narrative over the past two years has centred on adoption velocity: how quickly enterprises can bolt generative AI or agentic systems onto existing workflows. Alteryx’s data suggests the more pressing question for 2026 is not how fast organisations can deploy AI, but whether they have built the underlying data foundations and accountability structures to actually trust what it outputs.
Why it matters for the region
Singapore’s caution here is worth watching precisely because of its outlier status. As a market often seen as further along the AI maturity curve than its Southeast Asian neighbours, its analysts’ reluctance to cede control signals that the “trust gap” between AI capability and AI governance may not simply close with more advanced tooling or bigger budgets.
Also Read: Singapore’s AI tools are ready. Its workforce isn’t
If anything, the report suggests that as agentic systems become more capable and more embedded in day-to-day business decisions, the human-in-the-loop instinct may harden rather than fade, with organisations that invest early in data governance and clear decision ownership best placed to convert AI adoption into genuine business results.
The post Singapore’s data analysts trust AI to work, not to think appeared first on e27.
