EY projects that the integration of generative AI (GenAI) into business could significantly boost the global GDP by a staggering US$1.7 trillion to US$3.4 trillion in the next decade.
While the projection sets a positive tone for AI's potential impact, 71% of Asia-Pacific CEOs cite uncertainty around the technology as a challenge in developing and implementing an AI strategy.
AI integration challenges
According to EY's Greater China AI Leader Chris Leung, businesses are confronted with numerous challenges when integrating AI into their operations.
He said a common hurdle is the lack of in-house expertise. Finding skilled AI professionals with technical expertise and compliance knowledge is a significant challenge, as demand far exceeds supply.
"This necessitates significant investment in training or hiring external experts, and the temptation is to use existing talents and retrain, particularly for generative AI (GenAI). A CIO needs to balance the potential game-changing disruption of AI against this investment," said Leung.
Aside from talent issues, data privacy and security concerns remain challenging for businesses attempting to integrate AI into their operations.
He explained: "With stringent regulations in many APAC countries, AI systems must comply with local data protection requirements. For instance, China and Japan's regulations often prohibit cross-border data sharing. This includes safeguarding sensitive information from breaches, ensuring that AI models are trained on secure, high-quality data, and addressing data ownership issues before model deployment. Otherwise, regulatory intervention and fines are likely in the future."
Additionally, Leung said that gaining customer acceptance and trust can be difficult, citing the 2024 EY Australian AI Sentiment Report, which revealed that only 35% believe the benefits of AI outweigh the potential negatives.
It is no use building expensive models if customers opt out of using them or don't take their recommendations.
"Financial institutions need to take their customers on the AI journey, define responsible AI practices to promote the trust of their AI usage, and properly disclose the use of AI in their customer engagement. It is no use building expensive models if customers opt out of using them or don't take their recommendations," he said.
Independent AI implementation
Deploying AI into business functions can be complicated, but Chris Barford, EY Hong Kong's Financial Services Consulting AI & Data Lead, believes companies can implement AI themselves.
"Insurers with fewer regulatory constraints are experimenting more than banks in APAC. In addition, certain smaller players – particularly digitally native banks – see AI and GenAI as a chance to leapfrog the competition. However, with the very latest GenAI tech, almost no firms can afford to develop their foundation models. They need to use large language models, which can be expensive to run and complicated to customise."
However, he highlighted that they must consider factors crucial for AI implementation, such as experience and expertise.
In terms of experience, he explained that companies should be able to properly manage innovation by "identifying opportunities for AI applications and scaling successful innovations quickly" to succeed in a rapidly evolving landscape.
Moreover, he highlights the importance of change management for smoothly transitioning to AI-driven operations and vendor management to balance the value of the new AI technologies with potential risks such as intellectual property lawsuits.
Regarding expertise, Barford underscores compliance, regulatory knowledge, and AI model selection and evaluation.
"By prioritising these areas, CIOs can build a strong foundation for successful AI implementation and leverage its capabilities to drive innovation and growth," Barford said.
AI strategy and business objectives
Aligning AI strategies with broader business objectives is a critical step in avoiding underinvestment and misaligned priorities, as suggested by Leung.
"Leading players often have dedicated AI strategies that define their approaches to deploying AI for everyday productivity and identify key transformation pillars. These pillars enable strategic changes, such as creating new experiences and building new products, which can become future competitive advantages," he explained.
Further, he said that a partnership strategy ensures "a clear approach to third-party partnerships and in-house talent development". Understanding the ecosystem of AI technologies helps companies develop a plan to decide what to procure and build in-house.
Finally, Leung believes that cost optimisation of ongoing AI application operations is key to capturing ROI from generative AI applications.
"Identifying the right technology framework tailored to specific volume, performance, and accuracy requirements is critical. CIOs should maintain strategic flexibility to switch between different technologies, ensuring that cost-performance remains competitive with the latest market standards," Leung explained.
Building confidence about GenAI's value
According to Barford, financial markets see GenAI as a transformative force, potentially boosting knowledge worker productivity by 30-40%—similar to how LEAN revolutionised manufacturing in the 1960s. However, in reality, businesses are facing early-stage setbacks.
"We have likely already fallen into the trough of disillusionment, as the early promise of customer tech like ChatGPT has not easily transitioned into the workplace, particularly not in regulated industries like financial services," Barford explained.
Despite AI technologies' uncertainties, Barford listed ways organisations can reduce uncertainty and build confidence in them.
First, he said it is vital to start small to harness AI's full potential in the workplace.
Adopting a mindset of continuous improvement and innovation encourages organisations to evaluate and refine their GenAI initiatives regularly.
"Implementing small-scale pilot projects allows organisations to test GenAI applications in a controlled environment. These projects can demonstrate tangible benefits and provide valuable insights into potential challenges. Adopting a mindset of continuous improvement and innovation encourages organisations to evaluate and refine their GenAI initiatives regularly," he said.
Moreover, he reminds companies to establish a robust governance and ethics framework for responsible and compliant AI use.
"This includes defining ethical guidelines and setting up oversight mechanisms to monitor AI usage and manage long-term performance, systematically identifying risks and areas for improvement," he said.
He added that maximising AI in business also involves partnerships with technology providers, third-party service providers, and academics.
Barford said: "Given the rapid advancements in AI fields, these partnerships help organisations focus on where to invest, which provides competitive advantage, and decide what can be leveraged from the ecosystem. These collaborations can help organisations navigate the complexities of GenAI and enhance their implementation strategies. An organisation that attempts to do this all by itself will likely over-invest and under-deliver, as the technology is complex, and the required hardware is expensive to run."
Maximising ROI from AI
To maximise ROI from AI, businesses must identify the most impactful areas for AI integration. Leung said companies must identify the most impactful areas for AI integration through these strategic steps:
- Assess business goals and challenges
- Evaluate the potential for automation
- Review data and workflow integration complexity
- Establish a clear human-in-control strategy
"Given current technological constraints, AI, particularly GenAI, will not be 100% accurate. It is critical to define how to incorporate "human-in-control" to prevent and/or mitigate errors or improper decisions made by AI, maintaining high accuracy and reliability," Leung explained regarding the last one.
Technologies to invest in
Leung said there are valuable AI-related technologies organisations must invest in 2025 and beyond.
The first is Agentic AI. With its ability to perform complex tasks, optimise operations, and collaborate with other AI agents, this technology represents a significant leap forward in AI technology.
Next is integrating traditional and generative AI so businesses can create a comprehensive AI platform that enables the organisation to deploy the most appropriate technologies for different business challenges.
He said companies should watch for China's AI technology: "For instance, Alibaba released a new version of its Qwen chatbot, QwQ, with "reasoning" capabilities. Chinese AI startup DeepSeek recently released an open-source model, DeepSeek-R1, which achieved technological breakthroughs in reasoning capabilities at a lower price than was thought possible. Additionally, China's advancements in facial recognition technology and autonomous vehicles highlight its leadership in AI applications. By exploring Chinese AI technologies, CIOs can diversify their technological portfolio and potentially achieve better cost-performance for their AI applications."
Avoiding the AI underinvesting trap
Underinvestment in AI, said Leung, can lead to significant drawbacks such as outdated or inadequate AI systems and misuse of AI models, which can lead to reputation risks, data breaches, and cyberattacks.
Leung said businesses must integrate AI or "digital workers" into their workflows to stay competitive. "These digital workers can handle repetitive and time-consuming tasks, allowing human employees to focus on more strategic and creative activities," he added.
He added that appropriately investing in AI can increase productivity, job satisfaction, and innovation. "Businesses can thus achieve greater efficiency, accuracy, and scalability, leading to a stronger competitive edge in the market," Leung said.
For Leung, the sooner organisations experiment with and adopt emerging technologies, the better positioned they will be to evolve and become future-ready.
