Artificial intelligence (AI) adoption among Singapore enterprises is accelerating. According to IMDA’s Singapore Digital Economy Report 2026, the enterprise AI adoption rate rose from 14.7% in 2024 to 23.5% in 2025. Adoption among small and medium-sized enterprises (SMEs) increased from 14.5% to 23.4%, while adoption among larger enterprises rose from 62.5% to 70.4%.
These figures show that AI is no longer limited to large technology companies or multinational corporations. Increasingly, SMEs are using AI for content creation, customer service, process automation, document management, data analysis and internal operations.
But higher adoption does not mean that businesses have mature AI capabilities. IMDA’s data also shows that more than two-thirds of surveyed workers recognised a need to upskill or reskill in AI, while only about 37% had attended AI-related training during the preceding 12 months. The pace of business adoption is moving faster than the development of internal training, controls and governance systems.
The next question for businesses is therefore not simply whether to use AI. It is: Can employees use AI correctly? Can the business control the risks? Can management explain how AI creates value? Can the company demonstrate that important decisions were properly reviewed?
Singapore’s digital economy is now a major part of the wider economy
IMDA reported that Singapore’s Digital Economy reached S$144.1 billion in nominal value added in 2025, representing 19.3% of GDP, up from 18.8% in 2024. The Digital Economy grew by 5.6% year on year, ahead of nominal GDP growth of 3.1%.
Digitalisation outside the Information and Communications sector contributed approximately S$97.0 billion, representing the larger share of the Digital Economy. The Information and Communications sector contributed approximately S$47.1 billion. This means that Singapore’s Digital Economy is not being driven only by technology companies. Digitalisation in finance, wholesale trade, manufacturing, professional services and other traditional sectors is equally important.
The number of tech professionals increased by 3.8% in 2025 to 222,200. Tech employment in non-Information and Communications sectors grew by 5.3%, compared with 1.8% in the Information and Communications sector. Technology talent demand is therefore spreading across traditional industries rather than remaining concentrated within technology companies.
In 2025, approximately 96.4% of enterprises adopted at least one of six digital technologies: cybersecurity, cloud computing, e-payments, e-commerce, data analytics and AI.
Digitalisation is no longer an optional technology project. It is part of the operating infrastructure of a modern business. AI is becoming the next layer of productivity and decision support.
Why AI adoption is accelerating
Businesses are adopting AI in different ways. Many SMEs are not starting with complex autonomous systems or fully customised foundation models. They are beginning with specific, repetitive and measurable tasks, such as:
•drafting, editing and translating marketing content;
•handling routine customer questions;
•organising meeting notes, document summaries and internal searches;
•preparing quotations, reports and management materials;
•performing basic data organisation, classification and analysis;
•automating parts of administrative and operational workflows; and
•helping employees retrieve and synthesise information.
IMDA reported that close to nine in ten AI-adopting enterprises use commercial off-the-shelf AI tools, while 28.3% customise those tools for enterprise-specific needs.
The business case is generally practical rather than ideological. Among AI-adopting enterprises, 87.6% cited productivity and process improvement as a key benefit, while 44% reported cost reduction and resource optimisation.
SMEs that adopted AI-enabled, pre-approved digital solutions through the Productivity Solutions Grant reported average cost savings of 49.5% in 2025. This figure is based on self-declared outcomes from PSG applicants and should not be treated as a universal result that every business will achieve after adopting AI.
Almost one in four SMEs now use AI
The increase in SME adoption from 14.5% to 23.4% is one of the most significant developments in the report.
Roughly one in four SMEs has started to use AI. However, adoption rates do not measure depth of deployment. One company may use AI only for drafting and summarisation. Another may connect AI to customer service, financial analysis, sales operations or an internal knowledge system. The requirements for data security, workforce capability and management controls are very different.
Businesses should distinguish among three stages of adoption.
The first is individual experimentation. Employees use AI tools for isolated tasks without a common company policy.
The second is departmental use. A team incorporates AI into a recurring workflow and develops stable use cases.
The third is enterprise deployment. AI becomes connected to company data, software systems, customer processes or important decisions. At this stage, the business needs clear access controls, vendor management, human review, record retention and accountability arrangements.
Moving from the first stage to the third requires more than buying additional software licences. Governance capability must develop at the same time.
Employees are using AI faster than they are being trained
AI adoption is not limited to companies. IMDA reported that approximately 85.7% of surveyed workers used AI at work in 2026, up from 77.6% in 2025. AI users generally reported improvements in productivity, work quality and problem-solving ability.
Employees are using AI across a broad range of tasks. Content creation remains the most common use case, followed by process automation, data exploration and analytics.
Capability, however, is uneven. Many employees can use AI for general tasks but may not be able to design complex instructions, evaluate outputs, identify bias, verify facts or protect sensitive information in higher-risk settings.
Approximately 68% of workers recognised a need to upskill or reskill in AI, while only around 36.7%, commonly rounded to 37%, had attended AI-related training in the preceding 12 months.
This creates a clear capability gap: employees have begun using AI, but businesses may not yet have matching training, approval and quality-control systems.
What untrained AI use can expose a business to
If a business encourages employees to use AI without defining boundaries and providing training, several risks may arise.
Data and confidentiality risk
Employees may not know which company materials can be entered into an external AI service. They may unintentionally upload client information, personal data, source code, contracts, financial information or trade secrets to an unauthorised platform.
Accuracy and professional-judgment risk
Employees may treat AI-generated material as fact and overlook source verification, calculation checks or professional judgment. AI-generated legal, tax, accounting, human-resources and financial content requires particular care and human review.
Intellectual-property risk
The company may be unable to establish the source of an output, the rights attached to input materials, the presence of third-party protected elements or the permitted scope of use. Before external publication, the business should conduct appropriate copyright, trademark, trade-secret and contractual checks.
Access and accountability risk
If employees use personal accounts, unauthorised plugins or unapproved connections to internal data, the company may not know who had access, who approved the use, who reviewed the result or who is responsible for remediation if an error occurs.
Operational and reputational risk
Unreviewed AI output that enters client communications, financial reports, marketing materials or management decisions may cause client losses, contractual disputes, regulatory issues and reputational damage.
These risks require particular attention in accounting, tax, legal, human-resources, corporate-advisory and financial-services businesses, where employees often handle highly sensitive and professional information.
Businesses should train employees not only to prompt, but also to judge
Basic AI training often focuses on prompts, product features and operating techniques. Professional businesses need a broader set of capabilities, including the ability to:
•decide whether a task is suitable for AI;
•identify whether the input contains protected or sensitive information;
•give an AI system accurate context, scope and constraints;
•check facts, calculations, citations and reasoning in the output;
•identify hallucinations, bias, omissions and unjustified certainty;
•recognise when a matter must be escalated to a manager or professional adviser;
•retain relevant prompts, inputs, outputs, modifications and approvals; and
•assess data-processing, retention, training and security terms when using an external AI provider.
AI capability is therefore not only tool-operating skill. It combines domain expertise, professional judgment, risk awareness and accountability.
Singapore is expanding national AI skills programmes
IMDA is expanding support for workforce capability through the National AI Impact Programme (NAIIP).
According to IMDA, over the next three years, 40,000 tech professionals will receive advanced AI skills training, including skills relevant to full-stack development and agentic AI. NAIIP will also train 100,000 non-tech professionals in domain-specific AI skills by 2029.
IMDA’s programme materials emphasise that NAIIP is not intended to separate professionals from their existing fields. Instead, it combines professional expertise with practical AI skills. Current pathways include technology, accountancy and legal practice.
This reflects a policy shift from encouraging businesses merely to purchase AI tools toward helping businesses and workers use AI effectively and responsibly in real work.
There are two implications for businesses.
First, companies should monitor government-supported training, skills-development and sector programmes and select options that match their workforce needs.
Second, businesses cannot outsource all training responsibility to the government or a vendor. Each company still needs internal rules and role-specific capability standards based on its own data, clients, workflows and risk profile.
A practical AI-management foundation for businesses
1. Maintain an inventory of AI tools and use cases
Record which departments use which AI tools and for what purposes. Classify use cases by risk, distinguishing the organisation of public information from the processing of internal information, client data, code and support for significant decisions.
2. Define permitted and prohibited inputs
State clearly which materials may not be entered directly into external AI tools. These may include personal data, client confidential information, trade secrets, unpublished financial data, source code, draft contracts and materials subject to third-party restrictions.
3. Establish human review and escalation rules
Define which outputs require review by a manager, professional adviser or data-protection lead. AI-generated material involving external publication, client advice, financial figures, legal conclusions, tax treatment, employment decisions or significant business judgments should not be used without appropriate review.
4. Integrate AI training into role-based capability frameworks
Different roles require different AI capabilities. Marketing teams need to understand copyright and brand risk. Finance teams need accuracy and audit trails. Legal teams need confidentiality and source controls. Developers need code security, licensing and software-supply-chain controls. Training should reflect the role rather than rely only on a generic course.
5. Retain necessary use and approval records
For important projects, retain information about the AI tool, input sources, outputs, human modifications, review comments and final approval. The purpose is not to create unnecessary administration. It is to ensure that the company can explain its decisions if a dispute, error or due-diligence request arises.
6. Review vendor and system risk regularly
Assess the AI provider’s terms on data retention, model training, cross-border transfers, access rights, service interruption, output responsibility and security controls. AI systems connected to internal data or authorised to perform actions should also be subject to access testing, logging and incident-response arrangements.
Higher adoption is only the beginning
Singapore’s enterprise AI adoption rate reaching 23.5% shows that AI is moving rapidly from experimentation into day-to-day business operations. The Digital Economy reaching S$144.1 billion and 19.3% of GDP confirms that digitalisation is now a central part of the wider economy.
Long-term value will not depend on which business owns the most tools. It will depend on whether the business can connect technology, people, data and governance.
Businesses should now pursue three priorities together: give the right employees role-relevant AI capabilities; keep sensitive data and professional judgment within controlled boundaries; and preserve explainable, traceable records of important AI use and decisions.
Mature AI adoption does not mean asking employees to use AI as much as possible. It means being able to answer clearly: What is AI being used for? Who is authorised to use it? Who reviews the result? How is the output verified? How does the company manage the resulting risks?
