Singapore’s developers are among the most enthusiastic adopters of AI in the world, but a growing body of evidence suggests the AI infrastructure underpinning that ambition is falling dangerously short.
A survey of 196 developers and tech leaders conducted at API Days Singapore in April by customer engagement platform Twilio found that 96 per cent of respondents already use AI tools in their daily workflows. Yet for many organisations, broad adoption has not translated into meaningful outcomes. The culprit, according to the findings, is a fractured AI infrastructure that cannot support the demands being placed upon it.
Nearly half of respondents — 46 per cent — identified constant context-switching between disjointed tools as the primary source of friction at work. Poor integration between platforms was flagged as the single biggest barrier to achieving effective synergy between AI and enterprise automation.
Over a third of those surveyed (35 per cent) reported struggling with tools that simply cannot communicate with one another, while 24 per cent said they were contending with siloed data spread across multiple disconnected systems. For businesses that have invested heavily in AI tooling, the drag created by weak AI infrastructure is quietly eroding those gains.
Leadership gap stalls AI at the pilot stage
The underlying cause of much of this fragmentation is a lack of strategic direction from the top. Fewer than 30 per cent of respondents said their organisations had a clear strategic vision for AI deployment. Among founders and startup leaders, 41 per cent admitted they were still testing AI tools without a formal framework to guide adoption.
When individual teams are left to select their own tools without a unified plan, the consequences compound quickly. Forty-one per cent of respondents said their data was now scattered across too many disconnected systems — a direct result of decentralised decision-making.
The consequences for delivery are stark. Nearly a third (31 per cent) of organisations without a formal AI strategy struggle to move initiatives into production. By contrast, only three per cent of organisations with a structured roadmap face the same problem. Robust AI infrastructure, combined with strategic oversight, appears to be the differentiating factor.
Misaligned priorities between teams are accelerating tool sprawl. Sixty-one per cent of software engineers ranked API availability among the most important criteria when evaluating new tools. Only 36 per cent of product managers shared that view, suggesting product teams are more willing to prioritise out-of-the-box functionality over long-term interoperability.
Without top-level coordination, those differing preferences quietly fragment an organisation’s data architecture, making coherent AI infrastructure increasingly difficult to maintain.
The stakes rise as agentic AI arrives
The urgency to address these infrastructure gaps is intensifying. Nearly 40 per cent of respondents said they are already building autonomous AI agents, while 25 per cent are integrating Voice AI to handle complex workflows. These systems — capable of scheduling meetings, processing refunds, and executing multi-step tasks — demand a level of cross-platform reliability that fragmented infrastructure simply cannot provide.
“Running next-generation models on fragmented legacy architecture is becoming a liability in today’s agentic ecosystem,” said Michelle Duke, Senior Developer Evangelist at Twilio. “The missing link is the connective tissue between these isolated systems.”
The post Singapore’s AI infrastructure gap is trapping businesses in pilot purgatory appeared first on e27.