ST Explainer: What contributes to success in AI?
Sign up now: Get ST's newsletters delivered to your inbox
In his Budget 2026 speech on Feb 12, Prime Minister Lawrence Wong laid out plans to ramp up national support to help businesses adopt AI.
ST PHOTO: LIM YAOHUI
- Singapore actively promotes AI adoption with tax incentives, grants, and free tools, offering substantial government support to businesses for consultancy and generative AI initiatives.
- AI learns from data for decision-making, whereas automation executes predefined tasks. Companies must identify business problems first and ensure clean, accurate data is foundational for effective AI.
- Small businesses can start with quick, contained uses which builds trust, enabling companies to scale AI as a core business function.
AI generated
SINGAPORE - Singapore is making a concerted push for artificial intelligence (AI) by dangling tax incentives, grants and free subscriptions
In his Budget 2026 speech national support
But companies cannot jump on the AI bandwagon if they have not already automated their processes. Automation and AI are often used interchangeably, but they are different.
The Straits Times explains why AI implementations must be built on automation and why clean data is needed for companies to glean business insights and benefit from using AI.
Q: What are the differences between automation and AI?
Automation is the use of technology or programs to perform tasks such as invoice generation or inventory updating with minimal or reduced human intervention by following a set of predefined rules.
In contrast, AI is a branch of science that can learn from data patterns to perform more complex tasks, such as recommending products to customers and improving over time.
Newer forms of AI, known as generative AI, can even allow users to create content, images or video by instructing the technology in natural language.
Mr Remus Lim, senior vice-president of the Asia-Pacific and Japan at US-headquartered AI and data platform company Cloudera, said enterprises need to automate first to remove manual effort and then add AI for intelligence and context.
Q: What are the prerequisites for AI?
First, define the business problem, say, inefficiencies of manual tasks or errors.
A March 2025 report by US tax consultancy Grant Thorton cited examples such as data entry and processing, inventory monitoring and customer support.
Second, perform a data audit to ensure it is accurate, complete and secure. AI’s effectiveness depends entirely on the quality of the data it uses.
A December 2024 report by US-based tech consultancy Sei said: “Identify where your data is stored, assess its quality, and determine if it is organised in a way that AI models can easily process.”
Data should be in standardised fields, tables or spreadsheets, which are easy for AI models to read or interpret. There should be no duplicates.
Q: What is the typical AI investment we are looking at?
Market estimates by Singapore-based AI training firm Pertama Partners put the number down to between $50,000 and $100,000 for businesses with fewer than 100 employees. This is for projects such as chatbots for customer service, or predicting if potential customers will likely make a purchase.
For spending between $100,000 and $200,000, companies can explore AI for the predictive maintenance of machinery, or to optimise their inventory. Estimates include initial technology set-up, staff training, as well as hosting and maintenance costs for the first year.
To defray the cost, local firms can tap existing grants. Under the Enterprise Compute Initiative (ECI), for instance, the Government covers 70 per cent of consulting cost, capped at $105,000 per company. Meanwhile, cloud service partners that have partnered the Government under ECI will offer companies cloud credits, training and AI tools.
Under another initiative SMEs Go Digital, companies can adopt pre-approved generative AI tools with up to 50 per cent grant support.
Q: What are the quick-wins companies can first shoot for?
Cloudera’s Mr Lim cited simple projects such as internal knowledge assistants to help employees find information, automatic summaries of customer interactions, anomaly detection that flags issues early, and predictive alerts that replace static reporting.
“Early chatbots relied on scripted responses and often frustrated customers because they could not understand context or intent. Modern AI assistants are different. When grounded in enterprise data, they can interpret requests, personalise responses, and support human agents during conversations,” he said.
“The goal is not to replace people, but to remove repetitive interactions so staff can focus on situations requiring empathy and judgment. The most effective customer service models combine automation and human expertise. AI handles routine queries instantly, assists agents with recommendations and summaries, and escalates complex cases to people.
“Each of these improves efficiency while proving the reliability of the underlying data,” added Mr Lim.
“In practice, successful organisations do not start by buying more technology. They begin by organising and governing their data so the same foundation supports analytics, machine learning and operational applications. AI then acts as a multiplier on existing investments,” he said.
Q: How do firms scale to reap measurable business results?
According to a report on AI adoption
Such businesses have redesigned their work processes to embed use of AI in areas like product development, supply chain management and customer engagement.
For example, instead of just deploying a simple chatbot for customer service, a manufacturing firm with an AI system that monitors and sends alerts on which machines require maintenance, and orders the parts needed for maintenance and repair, is one example of AI use that involves an entire business workflow.
