What I Learned at Fortune Brainstorm AI Singapore 2025

I spent last week in Singapore at Fortune Brainstorm AI Singapore, our second time hosting this event in the thriving city-state. Here are some of the key thoughts and impressions I took away from the conference:

The Pace of AI Adoption is Equally Fast Everywhere

Unlike previous technology waves where Asian companies lagged behind the U.S., Europe, and China, AI adoption is charging ahead at an equally ambitious pace across regions.

Everyone Wants AI Agents, But Few Are Using Them Yet

AI agents from OpenAI, Google, Anthropic, and others are here, but adoption trails hype globally. Agents are inherently higher risk than predictive or generative AI that simply produces content, and they are often unreliable at present. Some methods to improve reliability, such as deploying multiple agents with specific tasks and cross-checking, are expensive. Vivek Luthra, Accenture’s Asia-Pacific data and AI lead, noted most businesses currently use AI to assist human workers or as advisors for decision support, rather than automating entire workflows. Luthra predicts by 2028, one-third of large companies will deploy AI agents, automating about 15% of day-to-day workflows. This shift will be driven by falling costs, more capable models, and workflow redesigns to leverage agentic AI.

AI’s Impact on the Job Market is Not Yet Clear

Pei Ying Chua, LinkedIn’s APAC head economist, shared that despite anecdotal reports of young graduates struggling to find work, LinkedIn’s data on open roles does not yet show significant evidence. However, coders face increased competition, requiring more applications per job. Madhu Kurup (Indeed) and Sun Sun Lim (Singapore Management University) emphasized the growing importance of AI skills—like prompting and building AI agents—alongside human soft skills such as flexibility, resilience, and critical thinking. Jess O’Reilly from Workday predicts AI will encourage companies to adopt dynamic, team-based organizational structures resembling an “internal gig economy,” moving away from traditional vertical hierarchies.

Infrastructure is Destiny

Access to AI infrastructure is critical, even for countries that do not build their own models. Running AI models (“inference”) requires significant AI chip capacity. Building data centers demands large energy investments. Rangu Salgame, CEO of Princeton Digital Group, noted that fossil fuels, especially natural gas, will likely power Asia’s near-term data center expansion, posing climate challenges. However, AI data centers could accelerate renewable energy development in the medium term, such as solar and offshore wind.

Sovereign AI Matters, But Delivering It Is Challenging

Governments want sovereign AI to avoid dependence on U.S. and Chinese solutions, but achieving it is difficult. Besides the high costs of data centers and power infrastructure, training AI models to understand local languages and cultural nuances requires painstaking data curation. Kasima Tharnpipitchai, head of AI strategy at SCB 10X, which is building a Thai language large language model (LLM), stressed there are no shortcuts—“it really is just effort. It’s almost brute force.”

Embodied AI is China’s Big Strength

While U.S. and China appear evenly matched in AI model capabilities, China leads in “embodied AI” — AI integrated into physical devices like robotaxis and humanoid robots. Rui Ma, founder of Tech Buzz China, highlighted China’s control over the robotics supply chain and rapid progress in affordable, practical robots for factories and general use. A humanoid robot named Terri, using software from Hong Kong startup Auki Labs and hardware from Chinese company Unitree, impressed conference attendees.

There is a Middle Path Between the U.S. and China

Singapore continues to position itself as a bridge between the two superpowers. Digital Minister Josephine Teo shared that Singapore recently hosted AI safety researchers from the U.S., China, and elsewhere, culminating in the “Singapore Consensus”—an agreement emphasizing AI reliability, security, and alignment with human values.

Frequently Asked Questions (FAQ)

AI Agent Adoption and Reliability

Q: What is the current adoption rate of AI agents in businesses? A: While there is significant hype around AI agents, adoption currently trails behind. Most businesses are using AI for human augmentation or decision support rather than full workflow automation. Q: What are the main challenges to widespread AI agent adoption? A: AI agents are currently considered higher risk and often unreliable. Methods to improve their reliability, like using multiple agents for cross-checking, can also be expensive. Q: When is widespread AI agent adoption predicted? A: Vivek Luthra predicts that by 2028, one-third of large companies will deploy AI agents, automating approximately 15% of daily workflows, driven by falling costs, improved models, and workflow redesigns.

AI's Impact on the Job Market

Q: Has AI significantly impacted the job market for recent graduates? A: LinkedIn data does not yet show significant evidence of widespread negative impact on job availability for graduates, although coders are facing increased competition. Q: What skills are becoming more important due to AI? A: Skills like prompting and building AI agents are growing in importance, alongside essential human soft skills such as flexibility, resilience, and critical thinking. Q: How might AI change organizational structures? A: AI is predicted to encourage a shift towards dynamic, team-based organizational structures, similar to an "internal gig economy," moving away from traditional hierarchies.

AI Infrastructure and Development

Q: Why is AI infrastructure crucial, even for countries not building their own models? A: Running AI models, especially for inference, requires substantial AI chip capacity and significant energy investments for data centers. Q: What are the climate challenges associated with AI data centers in Asia? A: Near-term expansion is likely to be powered by fossil fuels, particularly natural gas, posing climate challenges. However, this could also accelerate renewable energy development in the medium term. Q: What are the difficulties in achieving "Sovereign AI"? A: Challenges include the high costs of data centers and power infrastructure, as well as the painstaking data curation needed to train AI models on local languages and cultural nuances.

Global AI Landscape and Strengths

Q: Which countries are leading in specific areas of AI development? A: China is noted for its strength in "embodied AI," integrating AI into physical devices like robotaxis and humanoid robots, due to its control over the robotics supply chain. Q: What is the significance of the "Singapore Consensus"? A: It represents an agreement among AI safety researchers from various countries, emphasizing AI reliability, security, and alignment with human values, positioning Singapore as a bridge between global AI powers.

Crypto Market AI's Take

The insights from Fortune Brainstorm AI Singapore highlight a critical juncture in AI adoption, particularly relevant to the financial sector and cryptocurrency markets. The rapid, yet uneven, pace of AI adoption across regions underscores the need for robust infrastructure and accessible knowledge. At Crypto Market AI, we focus on bridging this gap by providing accessible, AI-driven tools for cryptocurrency analysis and trading. Our platform leverages advanced AI agents and machine learning models to navigate the complexities of the crypto market, offering users real-time insights and automated trading strategies. We believe in empowering individuals with the tools to understand and participate in the evolving digital asset landscape, much like the emphasis on human-centric AI discussed at the conference. Explore our AI Agents to see how we're making sophisticated AI accessible for your crypto trading needs, and learn more about our approach to AI Data Analytics for Strategic Crypto Portfolios.

What I learned at this year's Fortune Brainstorm AI Singapore