While GenAI is becoming a vital tool in enterprises’ digital transformation, many are grappling with balancing its promise with practical execution. These barriers reflect the complexity of transitioning from experimental use cases to reliable AI solutions.
The latest to bear this out is an Economist Impact study, which showed that 91 per cent of Asean enterprises are using GenAI in at least one functional area of their operations. However, despite the widespread adoption, only 32 per cent of these businesses believe that their GenAI applications are ready for full-scale production.
The findings, released two days ago, was based on a survey of 1,100 technical executives, data engineers, data scientists and architects across the world.
For Asia-Pacific organisations, it found that key challenges impeding GenAI adoption were high implementation costs (40 per cent), a shortage of skilled talent (38 per cent), governance challenges (38 per cent), and issues with output quality (33 per cent).
However, Asean companies believe in the value of AI, with only 18 per cent of the region’s respondents believing that AI is overhyped, and 99 per cent expecting GenAI adoption across both internal and external use cases by 2027.
This concurs with worldwide trends, with global AI spend expected to reach US$1 trillion in the next few years, according to an earlier Goldman Sachs report.
For Asean, the Economist study reveals that 77 per cent of the region’s respondents see the technology as crucial to their long-term goals. Despite the momentum, 37 per cent believe investment across technical and non-technical domains is insufficient.
Seventy-four per cent of Asean businesses train their large language models (LLMs) with enterprise data, making them leaders in data intelligence.
In contrast, general-purpose LLMs without contextual enterprise data continue to be used by nearly half of data scientists (49 per cent) throughout Asia-Pacific. These general-purpose models often lack the quality, governance, and evaluative capabilities that enterprise-specific data can bring.
In terms of use cases, Asean organisations expect to use different models and tools in their agent systems, including open source and proprietary technologies, to drive better performance. Some 94 per cent expect to deploy open source AI models by 2027.
Notably, a minority or 20 per cent of Asean respondents say that their organisation’s data and AI governance are sufficient. Enterprises struggle with fragmented data estates, which complicate discovery, access permissions, data usage, audits and sharing. To comply with changing AI regulations, it is also essential to govern AI models and tools.
“It’s clear that AI is becoming an integral part of every business, and the technology is emerging as a critical driver of business growth,” said Cecily Ng, vice president and general manager for Asean at data and AI company Databricks, which sponsored the Economist Impact report.
“Yet enterprises remain cautious, balancing ambition with concerns around quality, cost, and implementation,” she added.
For many organisations, the real value of AI comes when the technology is unleashed on their own proprietary data to develop data intelligence, said Tamzin Booth, editorial director of Economist Impact.
“That data intelligence is even more valuable in an increasingly unpredictable world,” she said. “To drive the algorithm advantage they’re seeking, it’s clear enterprises must address significant challenges with producing high-quality outputs, identify ways to evaluate performance and governance with large AI models, and work out how to effectively connect AI to the workforce.”
