Singapore retailers are warned the greater risk isn’t bad AI governance — it’s inaction, as the technology moves faster than most businesses are ready for.

28 minutes ago

Photo by SiljeAO via Pexels

Centralised governance can support AI adoption across regional markets.

Singapore retailers are being urged to target artificial intelligence (AI) at specific business needs and set clearer rules around its use, as companies look to scale the technology without losing control of their data.

Retailers do not need to consolidate all their data before deploying AI, said James Wilson, a partner, technology consulting, advisory and head of consumer & retail at KPMG Singapore. Companies should instead identify the information needed to make a particular decision and build their AI application around that requirement.

"A local store does not need to know everything about all the data in the organisation," Wilson told the Retail Asia Summit 2026 in Singapore.

For retailers planning a three-month AI push, Wilson advised against beginning with data stored in Excel spreadsheets, which he said requires a heavy lift.

He divides AI applications into three categories, namely productivity, cost avoidance, and growth.

“Productivity covers faster reporting and store administration whilst cost avoidance covers predictive models that flag stock-outs”.

Growth, which Wilson called the ‘holy grail’, involves using customer data to create something new.

The three categories are not equally easy to fund, however.

“Productivity and cost-saving projects win budgets because they're tangible and you can measure the delta", he said.

Revenue-focused projects are more difficult to assess because they do not have an established baseline, making their potential returns harder to determine. Most firms are still at the proof-of-value stage for growth, he added.

Drawing on a quarterly KPMG survey of global AI leaders, Wilson said companies that are seeing value from AI track its true cost.

He also warned against allowing a long list of potential applications to delay action. Retailers should select use cases with a sound business case and return on investment before developing the modular architecture, data flows and AI governance needed to scale them.

For smaller retailers, one of the first hurdles is deciding where to begin.

Cindy Ngiam, director for retail at Enterprise Singapore (EnterpriseSG), said the agency has reorganised its Retail Industry Digital Plan (IDP) around AI to help more than 2,000 small and medium-sized retailers.

The plan now groups solutions according to business touchpoints: front of house, back of house and corporate functions.

“The change shifts the question put to retailers to where it hurts the most so they can start with solutions that resolve an immediate challenge,” Ngiam added.

Retailers commonly ask where they should begin and which providers they should work with, she said, even though many of the IDP's solution providers are companies they already encounter in their daily operations.

“The revamped plan is also weighted towards AI, with AI concierge, chatbot, demand forecasting and generative AI training amongst the offerings,” she added.

More than 75% of SME retailers had adopted entry-level digital solutions and 45% had adopted intermediate ones, according to the Infocomm Media Development Authority's 2025 retail survey. Adoption of more advanced technologies remained limited.

Ngiam said retailers in Singapore operate in a high-cost environment, making resource efficiency an important priority.
Technology can help retailers make better use of limited resources and prevent those constraints from limiting their growth plans, she said.

She cited Far East Flora's Clementi flagship as an example. Self-checkout reduced cashier workload by 30% to 40%, shortened queues and allowed employees to spend more time on customer service.

Generative AI also halved content production time, whilst AI-assisted digital marketing delivered more returns on the same spending, she added.

Clear AI guidelines

As retailers introduce AI into more parts of their businesses, they also need to set boundaries around how employees use the technology.

Pang Fu Wei, vice president of the Singapore Retailers Association and group managing director and CEO of Kim Hin International, said retailers should establish clear internal guidelines covering AI use, data and security risks.

Pang said the term “governance” can discourage SMEs because it can make AI adoption appear more complicated than it needs to be. He suggested treating it as a practical set of guidelines that explains how employees should use AI responsibly.

When his organisation introduced AI, the first steps were to establish those guidelines, assemble the necessary toolkit and train employees to use it responsibly.

That approach gives employees greater confidence to experiment whilst allowing the company to maintain controls as AI use expands.

“The greater risk is not doing anything at all because this technology is coming hard and fast,” Pang said.

For companies operating across the region, establishing those controls can be more complicated because laws and regulations vary between markets including Singapore, Malaysia, Hong Kong and China.

Pang said businesses should approach regional AI adoption from first principles, identifying what can remain consistent across markets and what may need to change as AI models evolve.

Data use is one area where companies need clear rules. Pang suggested a red, orange, and green framework, with customer information classified as a “hard no” and financial information permitted subject to appropriate safeguards

Companies should also determine which AI models employees can use and whether they should access them through personal accounts or company-managed business accounts.

Pang said his organisation opted for Claude Business at the time because of its approach to data use, while acknowledging that AI providers’ policies could change.

For retailers operating across several markets, Wilson said some AI controls should be managed centrally.

The centre should own governance, standards, architecture and measures of success, whilst local markets handle customisation, he said.

Data quality also needs attention before retailers feed information into AI systems.

Wilson warned that spreadsheet data can lack governance, with some files managed by a single person on their laptop. Feeding such information into AI models may not produce the right answer, he said.