6 signs Southeast Asia’s AI boom is entering its next phase

Southeast Asia’s first generative AI wave was largely about experimentation. Companies tried chatbots, employees tested copilots while startups rushed to build products on top of newly available foundation models. That phase is, however, starting to change. As revealed in a joint study by McKinsey, Singapore EDB and Tech in Asia, nearly half of Southeast Asian companies surveyed have now moved beyond AI pilots into scaling up AI technologies. That compares with a global average of 35%.

Nonetheless, this shift does not mean every AI project is delivering strong returns. In fact, many are not. Instead, businesses are asking harder questions about what AI actually improves, how it fits into existing systems and whether the cost is justified. As the region’s AI start-up scene is increasingly being reshaped by the boom in data centres and AI funding, Southeast Asia is moving beyond a period dominated by AI tools and demonstrations. The next phase will instead be shaped by deployment, infrastructure and measurable business value. Here are six signs that the transition is already underway.

---

---

1. AI is moving from pilots into everyday operations

More companies are moving beyond simply demonstrating that AI works. The McKinsey-EDB-Tech in Asia research found that 46% of surveyed Southeast Asian companies had moved beyond piloting AI initiatives into scaling them. Singapore was further ahead, with 56% of companies already scaling AI. Moreover, AI adoption is spreading beyond technology teams. Deloitte found that 46% of Southeast Asian CFOs reported pockets of AI use across their organisations, while another 12% reported extensive use. 

That changes where AI is being used. Instead of a standalone chatbot that employees occasionally test, companies are increasingly integrating the technology into customer service, software development and internal operations. A similar trend is appearing globally. McKinsey’s 2026 State of AI survey found that 44% of organisations were scaling AI across the enterprise, up from 38% a year earlier. More than half were using AI across at least three business functions.

For Southeast Asian companies, the next challenge is therefore less about finding a use for AI and more about making it reliable enough for everyday work. For example, banks can no longer treat AI systems that handle customer information as an informal experiment but instead as a mainstay of their processes. Likewise, manufacturers using AI to plan inventory have to ensure the system works with existing supply-chain data. Once AI becomes part of operations, integration and reliability matter as much as the quality of the model itself.

2. ROI is becoming more important than the demo

The second sign is a change in the questions executives are asking. Two years ago, showing that an AI system could write, summarise or answer questions was often enough to generate interest. Businesses now want to know whether it saves money, generates revenue or improves important business processes.

The issue is that measurable financial returns remain uneven. The McKinsey-EDB survey found that The McKinsey-EDB survey found that 60% of Southeast Asian respondents reported less than a 5% impact on EBIT from AI, while 18% reported no financial impact at all. That gap between enthusiasm and company-wide returns is likely to shape the next stage of adoption. Businesses are likely to become less interested in how many AI tools have been launched and more interested in whether a process becomes faster, cheaper or more profitable.

The next phase of enterprise AI may therefore be less about adding AI to an existing process and more about redesigning the process around it.

3. AI agents are moving into real workflows

The rise of AI agents is another sign the market is becoming more operational. Unlike a conventional chatbot that responds to a question, an AI agent can carry out a series of tasks towards a goal. That might involve more complex tasks such as information gathering, updating systems, preparing documents or passing work between different applications.

While the technology is still in its early stages, adoption is growing. McKinsey found that 40% of respondents from large organisations, defined as those with more than US$1 billion in annual revenue, reported scaling AI agents in at least one function in 2026, up from 27% a year earlier. For Southeast Asian firms, the potential use of AI stretches across industries. Financial institutions can use agents for customer support and internal research. HR teams can automate parts of recruitment and employee administration. Retail businesses can use them across marketing and customer service, while manufacturers can apply AI agents to supply-chain and inventory work.

However, the wave of enthusiasm also comes with a reality check. ASEAN Digital Times’ analysis of Deloitte’s 2026 technology trends notes that only 11% of organisations globally have agents fully in production, despite much higher levels of experimentation. AI agents are therefore moving beyond the demo stage, but the winners will probably be companies that redesign useful workflows rather than simply adding an agent to a broken process.

4. Infrastructure and data are becoming part of the AI strategy

As more companies scale AI, the technology underneath it becomes increasingly important. After all, while running occasional experiments through a public AI service requires relatively little infrastructure. Running AI across thousands of employees, customer interactions or factory processes requires much more computing capacity, reliable data and well-connected systems. More than US$50 billion has been invested by hyperscalers such as AWS, Google and Microsoft in AI-ready data centres and cloud infrastructure across Southeast Asia. Key players such as Malaysia, Singapore and Indonesia are among the markets competing for this investment.

However, more data centres alone will not solve the enterprise problem. AI systems also need usable company data. Roughly one in ten executives in the McKinsey-EDB survey identified data quality and availability as a major barrier to adoption, while companies continue to struggle with connecting AI to older systems and workflows. Cost is becoming another consideration as some organisations reported limited AI use because of high operating costs.

That puts infrastructure decisions closer to the centre of AI strategy. Companies increasingly need to decide what runs in the cloud, what data can be shared with external models and which systems need to remain private. Data architecture may sound less exciting than a new AI application, but it is becoming one of the factors that determines whether those applications can scale.

5. The talent challenge is shifting towards the existing workforce

Southeast Asia still needs more AI specialists, but hiring engineers alone will not be enough. One in five executives surveyed by McKinsey, EDB and Tech in Asia described talent as the biggest obstacle to scaling AI and producing measurable results. As a result, governments are responding. Singapore plans to triple its pool of AI professionals to around 15,000 and reskill thousands of workers, while Malaysia has programmes aimed at developing more than 13,000 AI talents.

However, enterprise adoption increasingly depends on people who are not AI specialists but other employees from other sectors who are capable of balancing and integrating AI usage into their daily processes. For example, a finance employee needs to know when an AI-generated analysis is unreliable, while a marketer needs to understand how to use automation without losing control of the brand.

That makes upskilling the existing workforce just as important as competing for scarce technical talent. The shift is visible in the productivity data as McKinsey found that 80% of global respondents believed AI had already improved their individual productivity and roughly half said it had helped them develop new skills. Companies that build AI capabilities of ordinary employees may thus gain an advantage over companies that concentrate expertise inside small technology teams.

6. Investors will want proof that AI businesses are actually being used

The final sign of this impending next phase comes from the startup market. During the first wave of generative AI, companies could attract attention by showing what new possibilities new models brought about. Yet as the technology becomes easier to access, that is no longer much of a competitive advantage. Instead, investors are likely to ask more difficult questions about how regularly customers are adopting AI products as well as how the specific startup differentiates itself from others through valuable data or specialist knowledge. 

The funding market is already showing signs of greater selectivity. Southeast Asia recorded its lowest quarterly deal count in at least eight years during Q1 2026, even as AI startups working on areas such as enterprise automation and AI agents continued to attract capital.

Infrastructure has also become part of the investment story. A handful of large AI and data-centre transactions have pushed up regional funding totals even while the number of startup deals remains relatively weak.

This makes revenue and genuine usage more important. Startups that can show an enterprise customer saving hours of work every week have a clearer case than one demonstrating another general-purpose AI assistant. Furthermore, companies with proprietary data, deep industry knowledge or products embedded into customer workflows may also be harder to replace.

Southeast Asia’s AI boom is becoming more mature

Southeast Asia’s AI market is no longer defined only by experimentation and rapid adoption. The next phase will be shaped by whether companies can integrate AI into daily operations, improve productivity and show clear financial returns.

Infrastructure, workforce skills and data quality will matter as much as access to the latest models. At the same time, investors are likely to become more selective about startups that can demonstrate real usage, recurring revenue and advantages that are difficult to copy.

The region’s AI boom is therefore changing rather than simply slowing down. It is becoming more practical. The companies that stand out in the next few years may not be those with the most impressive demos, but those that make AI reliable enough to become part of everyday business.