From Generative to Agentic: The Next Phase of AI Adoption for MSMEs

Prepared by Rashaad Ali

4 Jun 2026

Credit photo:NST Online

AI adoption in a rapidly diverse region

AI adoption in Southeast Asia has now surpassed the global average, reflecting strong potential for digital transformation. Yet this headline masks deep variation across countries.

The region is highly heterogeneous, with significant differences in infrastructure, skills, and regulatory maturity shaping adoption patterns. The release of ChatGPT in 2022 accelerated generative AI uptake, particularly among Micro, Small, and Medium Enterprises (MSMEs), by lowering entry barriers to experimentation.

However, while generative AI has expanded access, the next stage of value creation is increasingly linked to agentic AI — especially for MSMEs seeking efficiency gains rather than content generation alone. Across markets, a consistent pattern is emerging: AI delivers the greatest impact when it complements human decision-making rather than replaces it.

Uneven readiness across Southeast Asia

Singapore remains the regional leader in AI readiness and adoption. Its National AI Strategy (NAIS) launched in 2019 laid the foundation for coordinated development, followed by NAIS 2.0 in 2023 and the National AI Research and Development (NAIRD) Plan, supported by over US$1 billion in investment. These initiatives have strengthened research capacity, talent development, and enterprise adoption.

In contrast, Myanmar and Lao PDR remain at early stages of national AI policy development, still in drafting phases. Without coordinated frameworks, MSMEs in these markets face additional barriers in accessing structured support and capability-building ecosystems.

Cambodia illustrates the challenges of early-stage digital transformation. While AI awareness is growing among MSMEs, adoption remains concentrated among larger and more digitally mature firms. A recent study found that nearly 60% of Cambodian manufacturing SMEs still operate without core digital business systems and rely heavily on paper-based processes, underscoring weak digital foundations that limit AI readiness.

The Philippines reflects a similar trajectory. Although large ICT and BPO firms are increasingly adopting AI, fewer than 15% of businesses were using AI as of 2025. Despite relatively high internet penetration, MSMEs continue to associate AI primarily with chatbots and generative assistants rather than broader operational applications.

Agentic AI: from assistance to autonomy

Generative AI, a tool that produces material based on prompts, has become the first major entry point for MSMEs, supporting content creation, customer engagement, and ideation. However, it remains fundamentally reactive, responding to prompts without executing full workflows.

By contrast, Agentic AI, an autonomous system that independently manages workflows to complete tasks, represents a shift toward systems capable of planning, executing, and adapting multi-step processes with minimal human oversight. Unlike rule-based automation tools, which follow predefined instructions, agentic AI can respond dynamically to changing conditions and interact with external systems to achieve defined objectives. This positions agentic AI as a workflow orchestration layer, enabling MSMEs to move from task-level support to operational automation.

At its core, agentic AI is defined by four capabilities:

• Autonomy: independent planning and execution of tasks

• Adaptability: real-time adjustment based on feedback and changing conditions

• Tool integration: interaction with external platforms and systems

• Workflow orchestration: end-to-end execution of multi-step processes
  

For MSMEs, these capabilities collectively reduce reliance on specialised personnel while improving operational efficiency and responsiveness.

In practical terms, this shift is already visible across key business functions, where agentic AI moves beyond isolated task support to coordinating entire workflows.

In operations, for example, agentic AI can integrate sales data, inventory systems, and supplier platforms to enable real-time stock monitoring and automated procurement decisions based on demand fluctuations. Rather than requiring manual checks and approvals at each stage, the system can continuously assess inventory levels, trigger reorders, and adjust purchasing decisions dynamically.

In customer service, agentic AI extends beyond chatbot responses to manage the full customer interaction lifecycle — from initial enquiry and triaging to scheduling, follow-ups, and resolution tracking. This reduces fragmentation across communication channels and minimises the need for manual handovers, allowing MSMEs to maintain consistent service quality with lean teams.

In finance and risk management, agentic AI can continuously analyse transaction data to flag anomalies, detect potential fraud risks, and generate alerts or recommended actions. Unlike static rule-based systems, it can adapt to evolving transaction patterns, improving accuracy over time while reducing the need for constant human monitoring.

What distinguishes these applications is not just automation, but coordination. By linking multiple tasks into a single, adaptive workflow, agentic AI allows MSMEs to shift from reactive, task-based operations to more proactive and system-driven decision-making. This is particularly valuable for small firms operating with limited manpower, as it enables them to achieve operational scale and consistency without proportionally increasing headcount.

A 2025 study of MSMEs in Jakarta, Bandung, and Surabaya demonstrates measurable gains from agentic AI adoption across customer service, marketing, and financial administration workflows. Service cycle times were reduced from 2.4 hours to 1.1 hours, reflecting faster response and resolution rates. Customer satisfaction scores increased from 68% to 82%, driven by more timely and consistent engagement. At the same time, administrative workloads declined by up to 40%, particularly for repetitive and process-driven tasks.

Taken together, these outcomes highlight a broader shift: agentic AI is not simply improving efficiency at the margins, but enabling MSMEs to reconfigure how work is organised — transforming fragmented processes into integrated, intelligent systems that can operate with greater speed, accuracy, and resilience.

Barriers and Translating to Action

MSMEs across Southeast Asia continue to face four key barriers: access and affordability, capabilities and skills, awareness and use-case relevance, and trust, safety, and governance. Translating the potential of agentic AI into practice requires addressing each of these constraints through gradual, human-centred adoption pathways.

1.) Access and affordability constraints Uneven digital infrastructure and high costs continue to limit adoption, particularly in Myanmar, Lao PDR, and Cambodia, where low internet penetration and infrastructure gaps constrain basic readiness for AI use. In more developed markets, affordability remains an issue—MSMEs in the Philippines still face high broadband costs and slower speeds compared to regional leaders such as Singapore and Thailand, increasing the cost of digitalisation.

Translating this into practice requires MSMEs to adopt low-cost, low-code, and pay-as-you-go AI tools that reduce upfront investment and complexity. Firms can start small by piloting AI on specific, low-risk pain points before scaling gradually. Public–private partnerships are also important in widening access to tools and infrastructure. For example, in August 2025, the Department of Trade and Industry partnered with Canva to provide MSMEs with training and accessible tools, lowering entry barriers and enabling experimentation with more advanced applications such as agentic AI.

2.) Skills gaps and limited digital readiness

Persistent skills gaps limit MSMEs’ ability to effectively use emerging technologies such as agentic AI. While digital literacy is foundational, it remains uneven across the region—around 61% of individuals aged 10–24 in Southeast Asia are not receiving formal digital skills education, raising concerns about long-term workforce readiness.

Beyond literacy, effective AI use requires cognitive and workplace skills such as critical thinking, communication, and sound judgment. In the Philippines, over 21% of senior high school graduates lack reading comprehension skills, increasing the risk of misinterpreting AI outputs or missing errors such as hallucinations.

Translating capability-building into practice requires applied, hands-on learning that enables MSMEs to interpret, evaluate, and act on AI outputs. This includes managing expectations around AI’s limits and embedding ethical practices such as fact-checking and recognising AI-generated content. Programmes such as the LimitlessBiz course offered by the US-ASEAN SME Academy can support this, especially when complemented with country-specific case studies and local language translations.

3.) Awareness gaps and use-case limitations

Many MSMEs still associate AI mainly with chatbots and content generation, limiting awareness of broader applications such as workflow automation and autonomous task execution. This narrow framing can lead to hesitation or premature rejection of the technology due to perceived complexity or limited relevance.

Translating awareness into practice requires reframing AI as a workflow transformation enabler. MSMEs can start small by targeting specific operational pain points—such as inventory management, customer response cycles, or financial monitoring—and piloting AI on low-risk tasks before scaling. This incremental approach helps demonstrate tangible value and builds familiarity with agentic AI capabilities.

4.) Trust, safety, and governance

Trust remains a major constraint due to limited visibility into how AI systems process data and generate outputs, alongside concerns over privacy, accountability, and reliability. Underdeveloped governance frameworks further limit confidence in scaling adoption.

Translating trust into practice requires embedding responsible use into daily operations, including human-in-the-loop oversight, careful data handling, and verification of AI-generated outputs. MSMEs must ensure sensitive information is not input into systems and comply with organisational confidentiality policies.

Structured training also plays a key role. Programmes such as the course offered by the US-ASEAN SME Academy help build awareness of ethical AI use, including fact-checking outputs and identifying AI-generated content. Complementing such initiatives with localised case studies and language adaptations further improves accessibility and relevance across diverse MSME contexts.

Conclusion Agentic AI is the next step in MSME digital transformation in Southeast Asia, but its value depends on gradual, targeted adoption rather than full-scale automation.

For MSMEs, a well-paced integration leads to more sustainable long-term outcomes, rather than hasty implementation that may result in operational strain. An integrated support structure that leverages the resources and expertise of government, industry, and academia also provides a strong foundation for MSMEs to thrive amid rapidly evolving AI technologies.

Ultimately, the most effective approach is human–AI complementarity, where agentic AI augments human judgment to improve efficiency, resilience, and competitiveness.

Acknowledgements

Case studies and observations in this article draw on data and interviews with academic and industry observers in Cambodia, and industry practitioners in the Philippines.

The views and recommendations expressed in this article published in May 2026 are solely of the author and do not necessarily reflect the views and position of the Tech for Good Institute.