Singapore's small and medium-sized enterprises (SMEs) are adopting artificial intelligence (AI) quickly. However, many have not set up the systems needed to keep making progress if their main AI leaders leave.
New research commissioned by Amazon Web Services (AWS) found that six in ten SMEs across Singapore's financial services, healthcare and manufacturing sectors would face significant or moderate disruption if the primary person responsible for AI left the organisation. Around one in ten said their AI initiatives would likely stop altogether.
This finding highlights a new risk for companies that want to make AI a regular part of their operations. While more businesses are using AI, accountability, oversight, and continuity still need to improve.
"Singapore's SMEs have moved decisively on AI. The next step is making that investment sustainable by integrating AI across functions so it shifts from a point solution to a core part of how the businesses run," says Priscilla Chong, managing director for Singapore at AWS.
The study, conducted by research advisory firm Strand Partners, surveyed 1,500 businesses in Singapore. It included a nationally representative sample of 750 businesses and deeper sector surveys covering financial services, manufacturing and healthcare. AWS said the findings represent the views of the businesses surveyed and should not be read as an assessment of Singapore's national AI strategy or government frameworks.
AI is already widely used in the sectors covered by the research. In financial services, 75% of SMEs surveyed said they use AI. The numbers were 61% in healthcare and 57% in manufacturing. However, fewer companies have reached advanced use, which the study defines as combining several AI models or tools, or building their own AI systems. Among those using AI, only 29% in financial services, 23% in manufacturing, and 16% in healthcare said they had reached this advanced stage.
See also: Why the plumbing should come before AI agents
This gap is important because it shows that many SMEs have moved past the early stages, but have not fully made AI part of their decision-making, review, and scaling processes. As a result, AI projects can be affected by staff changes, informal decisions, and uneven confidence among employees using these tools.
Accountability is where the issue stands out most. In all three sectors, at least half of SMEs say people either keep final control or share decisions with AI systems. Still, only about 30% of SMEs using AI said they have a clear person responsible for checking AI accuracy. In other cases, responsibility is shared among teams, given to the team that built the system, or decided as needed.
This lack of structure may be harder to manage as AI spreads beyond tech teams. The research found that IT remains the primary leader in AI projects, but other areas, such as marketing, human resources, and operations, are also starting projects. In all three sectors, fewer than 40% of businesses said they do not have a formal process for escalating AI outputs that employees are unsure about, indicating that the majority lack a clear channel for raising concerns.
See also: Kimi chatbot maker Moonshot AI valued at US$20 bil in Meituan-led round
"The opportunity is clear. Businesses that build clear, simple escalation routes, something as straightforward as a defined step for when a team member questions an AI-generated forecast before it reaches a client, will move faster and get more from their AI investments," says Chong.
The main obstacles vary by industry. In financial services, 38% of SMEs said getting internal approval and sign-off takes the most time when deploying AI. In healthcare, 30% said the same. In manufacturing, the biggest challenge was connecting new systems to existing workflows, mentioned by 37% of businesses.
These differences show why general AI advice may not be very useful for SMEs. About two-thirds of businesses that found industry-specific advice still had to change it significantly before it worked for them. Only 20% of healthcare SMEs, 17% of financial services SMEs, and 13% of manufacturing SMEs said the advice they found fit their needs directly.
For financial services firms, the main problem might be getting approval to use AI. For manufacturers, it could be linking new tools to current systems and workflows. In both cases, the challenge is to create enough structure for AI without making things slower.
"The companies getting this right tend to share a common approach: a dual strategy. One environment that's safe and open for experimentation, where employees across the business can build familiarity and confidence with AI tools. And another with strict guardrails for production, where proven use cases run at scale against real business problems," says Chong.
