AI is evolving at breakneck speed. Every week brings new capabilities, bold launches, and a fresh wave of funding. Yet, this momentum creates a practical problem for business leaders. With the landscape shifting so quickly, how do businesses decide where to invest, what to build, and which problems to solve first?
At NCS, we have been working with companies and governments across Asia Pacific (Apac) to accelerate their AI deployment. In our conversations with these organisations, we see three themes that we believe will shape how AI evolves and how enterprises extract value. The technology is changing, the supply environment is fragmenting and expanding, and demand is maturing toward resilience and returns. Understanding these three forces will help business leaders invest with clarity and deploy AI that delivers.
The shift to right-sized AI
The era of chasing the biggest model is giving way to using the right model for the right task. AI is moving beyond a one-size-fits-all approach. Models like ChatGPT-5 now route queries to the right model based on complexity: simple questions are handled by fast, lightweight models, reasoning tasks are escalated to specialised thinking models, and agentic models perform multi-step actions like web searches or analyses. This approach optimises speed, cost, and capability, choosing the right tool for the task rather than over relying on a single massive model.
Large language models (LLM) are no longer text-only. Modern LLMs can understand and generate text, images, video, and audio, combining multiple modalities to produce richer outputs. For example, AI can analyse website images, videos, and graphics to provide comprehensive answers, or generate real-time voice and multimedia content. This transition to multimodal intelligence enables more immersive and effective AI interactions.
For businesses, this unlocks workflows that mirror how people actually work. For instance, we helped the Singapore Ministry of Manpower's contact centre to develop a Generative AI tool that streamlined contact centre operations and improved customer experiences. The tool provides accurate real-time transcription, understands regional accents, summarises transcripts, and generates accurate responses to complex queries within seconds. With this tool, the average call handing time reduced by 12%; the average after-call-work was reduced by more than 50%; overall productivity improved by 6%.
See also: AI overtakes growth plans in Asia’s boardrooms for 2026: report
Earlier AI relied on retrieval-augmented generation (RAG), pulling answers from sources like a textbook. Today, models are evolving to reason, breaking down questions into sub-steps, gathering and verifying evidence, and producing considered responses. This shift toward agentic AI enables models to perform complex, multi-step reasoning rather than simple recall, marking a significant evolution in generative AI capabilities.
Neutral, flexible architectures will win the race
A few years ago, GPU scarcity dominated headlines and recent geopolitical tensions have further fragmented AI technology stacks. Developers have adapted, shifting focus from raw compute to outcomes. DeepMind, for example, optimised limited resources to deliver a highly capable model. New players – from AMD and startups to nations building sovereign AI infrastructure – are expanding capacity and reliability. The key shift is clear: AI success now depends on turning available compute into scalable, impactful capabilities, not simply acquiring more GPUs.
See also: Alibaba Cloud partners with AI Singapore on a lightweight Southeast Asia-focused LLM
Geopolitical tensions have created distinct Western and Eastern AI stacks. Globally operating companies must navigate this divide, which calls for a federated, neutral architecture that can leverage multiple stacks. This enables access to top models from both regions – for instance, U.S.-based Google Gemini, OpenAI, and Anthropic, alongside China-based DeepMind, Baidu, and Alibaba – maximising capability while maintaining flexibility across markets.
Another trend is the move from global to hyperlocal AI models. Large models like GPT-4 and GPT-5 often miss local linguistic and cultural nuances. Locally tailored models, such as Singapore’s Sea-Lion and Meralion initiatives, capture Southeast Asian languages, dialects, and social context. By reflecting local culture, these models reduce bias, increase trust, and provide more effective, regionally relevant AI solutions.
The new AI mandate: ROI, resilience, and trust
Over the past two years, organisations have moved beyond experimentation and hype. AI adoption is maturing, with a focus on return on investment and business value. An adjacent shift is from pursuing raw innovation speed to building digital resilience, which is an organisation’s capacity to withstand, adapt, and recover from disruptions across its digital ecosystem, ensuring consistent and secure operations.
As AI adoption accelerates, it is easy to mistake velocity for readiness. But in the real world, speed without stability leads to fragility. At NCS, we believe you can only run as fast as your digital resilience. AI alone doesn’t create impact – it needs a foundation that ensures continuity, trust, and scalability. Cybersecurity, data quality, and infrastructure must be robust before AI is deployed in mission-critical processes.
Resilience is proven to drive ROI – a recent NCS-IDC survey of over 870 organisations globally found that companies with high AI adoption and strong digital resilience readiness achieved outsized returns: up to 3.7 times ROI from AI investments – 1.6 times that of peers.
Finally, governance is evolving from principles to operational assurance. Early AI discussions focused on guidelines, but countries are now implementing actionable frameworks. Singapore’s AI Verify Foundation, Australia’s national AI initiatives, and the EU’s AI Act provide controls, metrics, and reporting mechanisms. Embedding these assurance layers builds trust and enables AI to become a reliable part of daily operations.
To stay ahead of the latest tech trends, click here for DigitalEdge Section
The next era of AI leadership
The pace of AI will only accelerate. For business leaders, success lies not in chasing the biggest models or latest hype, but in right-sizing investments, embracing multimodal and agentic capabilities, and building flexible architectures that span global and local ecosystems. Equally critical is digital resilience – the foundation that ensures continuity, trust, and scale. Without it, speed becomes fragility. Companies that strengthen their resilience, invest with clarity, and adopt neutral, scalable strategies will capture outsized ROI while navigating geopolitical and technological shifts. The mandate is clear: lead with resilience, and deploy AI that delivers measurable, lasting impact.
Ying Shaowei is the chief scientist at NCS
