New research by Twilio shows that Singapore’s AI adoption is widespread at 96%, but infrastructure issues such as fragmented tooling and poor integration significantly hinder AI productivity.
Holding back AI productivity
The survey found that the primary source of friction in developers’ daily workflows includes constant context-switching between disjointed tools (46%), incompatible tools (35%), and siloed data spread across multiple, disconnected systems (24%).
The survey indicates that leadership gaps and a lack of strategic infrastructure contribute to hurdles in AI deployment, with 41% of founders and startup leaders admitting they are still testing the waters without a formal AI adoption framework. Only about 30% of respondents say their companies have a clear strategic vision for AI deployment.
Nearly a third (31%) of organisations without a formal approach struggle to move AI initiatives into production. In contrast, just 3% of those with a clear plan face similar challenges.
The survey also found that misaligned priorities drive tool sprawl: 61% of software engineers rank API availability among the most important factors when evaluating tools, but only 36% of product managers (PMs) do, leading to fragmented decision-making and siloed tools that fail to integrate with the backend.
Need for a unified infrastructure
“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. Unlocking real AI productivity now requires a foundational infrastructure layer to integrate every endpoint.”
