In Southeast Asia, startups are no longer treating artificial intelligence as a feature or an experiment. AI and automation are becoming foundational. Generative tools are now being used to accelerate product development, streamline operations, and build user experiences that are contextually and culturally relevant. Investors are pushing for leaner teams that can achieve more with less, and startups are responding by embedding intelligent systems at the core of their businesses. The age of speculative AI hype has given way to a more operational reality where the question is no longer whether to use AI, but how effectively it is deployed.
Will inclusive innovation define Thailand’s AI in fintech push?
Agentic AI adoption is accelerating
Recent data from IDC Asia/Pacific highlights this shift. A 2025 survey shows that 42 percent of organisations in the region already use agentic AI, with another 44 percent planning to adopt it within the next 12 months. Among the most common applications are customer service automation, risk detection, and internal productivity enhancements. Over 70 percent of respondents believe that AI improves decision-making and output quality, and almost all agree that it boosts employee productivity. These figures suggest that AI is becoming a standard operating layer in Southeast Asia, not just a selling point for pitch decks.
Southeast Asia is building its own language models
The region’s diversity has also shaped how AI evolves in practice. Southeast Asia is home to hundreds of languages and dialects, which has prompted a wave of innovation in multilingual AI. Projects like SEA-LION and Sailor2 are developing large language models specifically trained on Southeast Asian languages, such as Thai, Malay, Burmese, Tagalog, and Vietnamese. SEA-LION, for example, uses instruction fine-tuning and region-specific data to enhance performance in local tasks, while Sailor2 supports multiple parameter sizes and is benchmarked for reasoning, reading comprehension, and contextual understanding across languages. Both are open-source and backed by researchers aiming to build more representative and accessible AI systems. In Indonesia, telco operator Indosat Ooredoo Hutchison partnered with tech giant GoTo to launch Sahabat-AI, a locally trained model designed to support Bahasa Indonesia and other national languages in customer applications. The emergence of such models marks a significant pivot from importing global tools to building foundational AI infrastructure tailored for regional needs.
AI use cases are expanding across sectors
Startups are already putting this infrastructure to use. In the Philippines, First Circle uses generative AI to streamline SME finance, helping reduce turnaround time for business credit decisions. In Indonesia, Xendit has adopted AI across its payment infrastructure to detect fraud and reduce downtime in digital transactions. Malaysian healthtech startup Nalagenetics uses AI in diagnostics and genomic analysis, while Thailand’s Botnoi Voice develops voice-based AI tailored for tonal languages like Thai. These are not experimental add-ons. They are embedded systems that drive commercial performance, reduce labour costs, and enhance end-user experience.
Infrastructure and policy are catching up
This acceleration is made possible by a broader ecosystem shift. Cloud infrastructure is expanding rapidly across the region, making computing power and storage more accessible to early-stage companies. Singapore remains the epicentre for AI infrastructure and regulation, while Malaysia, Vietnam, and Thailand are investing in research labs, data centres, and AI research hubs. Governments are also playing a more active role. Singapore has published frameworks focused on responsible AI, encouraging fairness, transparency, and explainability. Other countries are beginning to explore national AI strategies and have started integrating AI in public services, healthcare, and education. However, progress is uneven. A recent study of seven Southeast Asian countries noted that while the appetite for AI adoption is high, infrastructure gaps, limited talent, and unclear regulation continue to hamper more widespread deployment.
Talent remains a competitive bottleneck
Talent remains a bottleneck. While nations like Singapore and Malaysia have seen a 250 percent rise in AI proficiency, many others lag in producing machine learning engineers and researchers. Brain drain is a concern, with many skilled professionals migrating to better-funded ecosystems or joining multinational firms. Startups in markets like Vietnam, the Philippines, and Myanmar often struggle to compete for AI talent, which slows down development and adoption. In response, initiatives such as SEACrowd and other open-data projects are working to make regionally relevant datasets available for training and benchmarking, especially in underserved languages and modalities.
But several friction points remain
Despite the momentum, several challenges persist. Data quality is a recurring issue. Many startups use global foundation models that underperform in local contexts due to cultural bias or language misinterpretation. Lack of annotated data also limits fine-tuning and customisation. Meanwhile, integrating AI into legacy systems can be expensive, especially for SMEs. High upfront costs, compliance requirements, and interoperability issues are all common friction points. Moreover, with the regulatory landscape in flux, deploying generative AI across borders can be risky. Data sovereignty laws, liability frameworks, and AI ethics standards vary widely across Southeast Asia. Startups that fail to account for these differences risk fines, product bans, or loss of consumer trust.
Investors now expect AI to deliver outcomes
The implications for founders are significant. As AI tools become standard, differentiation will no longer come from whether a company uses AI, but how well it does so. Competitive edge will depend on the quality of training data, user experience, cultural nuance, model performance under constraints, and explainability. Investors are already asking startups to demonstrate cost savings, higher margins, and product scalability through automation. Those that can show impact in metrics—not just vision—are getting funded. This focus on execution is also shaping product strategy. Startups that can deliver AI solutions tailored to local markets are better positioned to scale across the region. Localisation is not just about translation, but about building systems that understand local buying behaviour, speech patterns, regulatory constraints, and trust markers.
Responsible deployment is no longer optional
One of the most notable shifts is in how startups treat regulatory readiness. It is no longer a task left to the legal department. Increasingly, compliance is becoming a product feature. For sectors like fintech, healthtech, and edtech, the ability to explain how models work, maintain audit trails, and prevent misuse is vital. Governments are watching, and so are customers. In such an environment, building responsible AI from day one is not a competitive disadvantage—it is a requirement.
The next 12 months will test SEA’s AI maturity
Looking ahead, several developments could shape the landscape over the next 12 to 18 months. The maturation of local language models like SEA-LION, Sailor2, and Sahabat-AI will determine how quickly startups can replace generic models with tailored ones. Regulation will play a defining role as governments codify standards for data usage, liability, and cross-border AI flows. Infrastructure investment—in data centres, regional compute capacity, and low-latency cloud services—will be crucial to support demand. And finally, startups will need to invest more in user testing, cultural understanding, and real-world robustness if they hope to build trust in AI-powered products.
Conclusion
Southeast Asia’s AI moment is no longer theoretical. Generative tools and automation are now being used not because they are trendy, but because they work. For startups, the path forward is clear: adopt AI where it drives results, localise it to meet real user needs, and scale it with responsibility. The bar has been raised. Success will belong to those who can meet it, not just with code, but with cultural fluency, operational discipline, and a sharp understanding of the markets they serve.