The initial AI frenzy has cooled, shifting from breathless hype to a more sober assessment of its capabilities and limitations. Yet, AI will remain a star player in 2025, promising to revolutionise industries and redefine our daily lives.
In the new year, businesses will begin to seriously consider automating workflows using GenAI. Figures from research firm IDC show that businesses in Asia-Pacific and Japan are leading the charge. By 2028, AI and GenAI adoption are projected to skyrocket to a US$110 billion industry.
Signs of AI maturation are already evident. The financial services sector has rapidly integrated AI for critical functions like fraud detection while industries such as healthcare are in earlier stages, focussing on building robust data infrastructure necessary to leverage AI for advanced diagnostics and patient care.
This immense market potential has triggered a frenzy among tech companies, all vying to deliver their AI-powered solutions. In today’s tech landscape, a tech product lacking an AI angle risks irrelevance, if not outright extinction.
The buzz phrase for 2025 is AI agent. Closely intertwined with this are critical themes such as AI governance and responsible AI; multi-modal and multi-model large language models (LLMs); and the evolving job landscape shaped by AI.
While other trends will undoubtedly emerge, these six are the standout AI trends in 2025.
1. AI agents become common: They represent a significant leap forward in AI applications, defined by their ability to act independently, adapt to changing circumstances, and revolutionise industries by driving productivity and efficiency.
These intelligent systems are designed to work alongside humans, automating routine tasks while escalating complex issues for human resolution. For instance, in a customer service workflow, an AI agent could handle initial inquiries, provide instant solutions for common problems, and seamlessly pass intricate cases to human representatives, ensuring both speed and accuracy.
This marks a transformative step in the human-AI relationship and signals the maturation of the technology, said John Roese, Dell Technologies’ global chief technology officer and chief AI officer, in his outlook for 2025.
Ying Shao Wei, chief scientist of IT services firm NCS, pointed out that AI is not just getting smarter, but also becoming more versatile. With agentic AI as a partner, businesses can transition from reactive strategies to proactive action.
In 2025, agentic AI promises to deep dive into enterprises, enhancing complex business workflows, optimising processes, and unlocking the return on investment for AI initiatives.
But what about small and medium businesses? They do not have the deep pockets and/or talent to deploy agentic AI, so the future less clear. What can they do to stay in the game? Will the authorities come to their rescue?
2. Moving beyond text to visuals: GenAI has evolved far beyond its initial text-to-text functionality. It now spans a diverse range of capabilities, generating high-fidelity images, animations, and videos instantly from text descriptions. Today’s foundation models, such as GPT-4, Gemini, and Llama2, can already do this.
Companies are already tapping into these advancements. For example, Coca-Cola has leveraged GenAI to create innovative Christmas advertisement, showcasing the commercial potential of multi-modal AI.
3. Focus on sovereign AI and data residency: As GenAI becomes more powerful, so does the importance of sovereign AI and data residency. Dell’s Roese said many governments seek to use AI for innovation. One way is to create localised AI ecosystems that are aligned with local culture, language and data security needs.
This trend sharply focusses a country’s ability to generate AI value and differentiation using its own infrastructure and data, fostering ecosystems that reflect local IP and priorities, he added.
An example is Singapore, which has developed a foundation model called Sea-Lion, which specifically understands and represents Southeast Asia’s linguistic and cultural diversity.
Data residency has become increasingly important today, as it serves as a fundamental component for large language models (LLMs). Consequently, governments are placing greater emphasis on data security and crafting new guidelines to protect sensitive information and ensure adherence to regulatory standards.
In 2025, there will be more countries developing their own foundation models in their own languages and governments enhancing regulations to protect sensitive data.
4. Responsible AI becomes moral imperative: AI hallucinations cannot be ironed out completely. There will continue to be inaccurate and false facts, invented citations or nonsensical outputs that do not correspond to prompts and reality.
Hence, as AI’s sophistication accelerates, ethical guidelines and regulations must continue to strengthen to address biases, ensure transparency, and mitigate potential risks.
Agentic AI, with its ability to autonomously take actions, raises concerns about accountability and potential misuse. NCS’ Ying said AI agents operating within business processes could make decisions that have significant real-world implications.
It is not inaccuracies or false facts that businesses must contend with, he added. “It’s essential to ensure that these agents align with organisational values and avoid unintended consequences.”
While significant progress in AI ethics and governance have been made at the LLM development stage, he pointed to downstream applications, which could have broad implications on companies’ reputations and have not yet received closer attention.
So, in 2025, expect businesses to focus efforts on responsible AI and governments to expand their oversight of AI governance to encompass the complex workflows and decision-making processes of AI agents.
5. Jobs will be displaced, jobs will be created: AI has impacted employment across sectors. However, new job opportunities will emerge with AI empowering workers to focus on higher-level tasks, creativity, and strategic thinking.
In 2025, Dell’s Roese believes that a whole new category of jobs will begin to emerge, such as AI ethicists and AI interaction designers who craft engaging customer experiences through AI.
Entirely new categories of technical jobs will also appear, he said. Think of engineers skilled in thermal plumbing able to boost overall performance of AI data centres by optimising heat and power and AI agent conductors who can integrate teams of people and AI agents together to increase productivity and efficiency.
6. The global AI race heats up: China, a major player in the AI race, is rapidly advancing its capabilities. Its centralised approach to AI development allows for swift deployment and strategic control. It challenges the Silicon Valley model, where investments go into individual companies that compete with each other for talent and projects.
In 2025, China’s approach to AI investment could potentially give it an edge in global AI competition. Expect major good-quality open source LLMs to emerge from China, whose 1.41 billion population provide an immense data pool for training.
Looking ahead
The six trends outlined above provide a glimpse into the dynamic and multifaceted nature of AI’s impact in 2025. Beyond these key trends, several other developments warrant close attention.
The emergence of smaller language models will make AI more accessible to smaller businesses and individuals, broadening its reach and impact. These models can be finely tuned for specific tasks or industries, delivering more accurate and relevant results.
Their efficiency allows for faster inference, enabling quicker response times and improved usability. They are likely to lead to a surge of innovative applications across various domains.
In cybersecurity, GenAI presents both significant opportunities and challenges. They will lower the entry barrier for cyber criminals, enabling them to create new malware and speed up their attacks. Phishing e-mail can be generated within 30 seconds using GenAI, said Jeremy Piazzala, consultancy firm EY’s lead partner for cybersecurity practice in the Asia-Pacific region.
The effective deployment of GenAI-powered cybersecurity solutions will be crucial in navigating the evolving threat landscape. In particular, Piazzala stressed cybersecurity awareness training needs to be updated with GenAI in mind. There is also an urgent need to inculcate a culture of extreme wariness of everything digital.
Equally, AI can be used to augment cybersecurity defenders, by helping to identify attacks and potential threats early. Piazzala expects GenAI cyber defence to be embedded in software, allowing end-users to triage and stop attacks. GenAI promises to provide answers much more quickly to tackle tomorrow’s attacks.
Yet another upcoming wave of AI revolution in 2025 is industrial AI, said Matthew Oostveen, chief technology officer for Asia-Pacific and Japan at data centre storage vendor Pure Storage.
Businesses are increasingly investing in in-house AI infrastructure, including expensive GPUs, to accelerate AI development and deployment. This trend, often referred to as industrial AI, will drive the adoption of advanced AI techniques like RAG (retrieval-augmented generation) and enable businesses to gain a competitive edge.
2025 promises to be a pivotal year for AI, as this transformative technology, often likened to the new electricity, begins to revolutionise work processes across sectors such as healthcare, finance, transportation and countless other domains.
Despite the challenges and uncertainties, AI holds immense potential to transform the world for the better. By fostering responsible AI development, addressing ethical concerns, and ensuring equitable access to AI’s benefits, its power can be harnessed to create a more prosperous and sustainable future for all.
