AI’s balancing act

‘AI is now utilised in more and more aspects of clients’ and our work. Business has no choice but to embrace it. Otherwise, the company will become uncompetitive and not meet clients’ expectations,’ says Andrew Lam, managing director of assurance at BDO Hong Kong, adding that AI is already woven into everything from client management systems to day-to-day audit and tax tools.

Security concerns

Given that financial services deal with highly sensitive customer information, cybersecurity and data privacy concerns were top-of-mind for 65% of participants in a joint survey by the Hong Kong Trade Development Council and PwC.

‘Security and privacy experts have raised alarms around potential vulnerabilities of emerging AI tools,’ says Andrew Wong, director, audit quality and professional practice at KPMG China. ‘These concerns are highly relevant to accounting firms like KPMG, where protecting client data is critically important.’

The rise of large language models and generative AI in accounting has introduced new complexities to security. The models, while powerful, can reveal sensitive information from their training data or generate hallucinated outputs that could compromise financial accuracy.

Mian Wong, director of advisory at Grant Thornton Hong Kong, stresses the importance of keeping a close eye on AI systems. ‘To avoid unfair generalisation by the model, accounting firms should include provision for AI model validation, data governance and algorithmic transparency to ensure that AI-generated insights remain accurate and unbiased,’ she says.

Comprehensive policies

Companies are tackling these challenges by putting comprehensive AI policies in place. KPMG has rolled out its Trusted AI principles to ensure employees are using AI safely and ethically while keeping client information secure. ‘We have been exercising caution in the use of emerging tools,’ says Wong at KPMG China.

First and foremost, businesses have to stay in line with local regulations. In Hong Kong, while there’s no single law governing AI use, the Office of the Privacy Commissioner for Personal Data has released guidelines on how to develop and use AI ethically. These guidelines give organisations, including accounting firms, specific advice on managing AI systems, evaluating risks and working with stakeholders.

Meanwhile, mainland China has taken a more targeted, sector-specific approach, implementing measures such as the Interim Measures for the Management of Generative Artificial Intelligence Services, which apply to any individual or organisation using generative AI technology to provide services to the public.

Guidelines to guardrails

‘Well-defined formal AI policies can provide a structured framework that allows firms to balance innovation with reliability in AI developments and usages,’ says Wong. She stresses that firms should prioritise policies covering privacy, intellectual property rights, human oversight, accountability and ethical AI governance.

For instance, Deloitte has implemented a Trust by Design framework that requires all AI applications to undergo rigorous testing before they can be used with clients. The firm mandates periodic reviews of AI models to ensure they stay accurate and free from bias, especially in areas like revenue recognition and fraud detection.

Even as AI becomes more commonplace, industry players are adamant that there’s no replacing human judgment when it comes to major financial decisions and audit findings.

‘Human oversight in AI-driven audits for critical evaluation, validation and correction of AI-generated outputs can reduce errors, enabling more informed and reliable financial reporting,’ says Angie Yim, associate director of Advisory at Grant Thornton Hong Kong.

PwC has gone a step further by setting up a Digital Intelligence Centre of Excellence that develops firm-wide standards for AI implementation and monitors compliance with these policies across all service lines.

Future adoption

Looking ahead, AI adoption is set to surge. KPMG’s research shows that nearly half (47%) of the 1,800 companies surveyed plan to make it a priority for financial reporting in the coming years. In supporting the ecosystem, the Hong Kong Monetary Authority has launched initiatives like the Generative Artificial Intelligence Sandbox where financial institutions can test AI applications in a safe and controlled environment.

The conversation around AI typically focuses on productivity gains, but Wong notes that it is just the tip of the iceberg. ‘It helps to reduce human resources on the more routine tasks and let the professionals focus on tasks that require professional judgements,’ says Lam. ‘Not only does this reduce the time and cost of the operations, but it also makes work more interesting and attractive for the employees, leading to better staff retention and utilisation.’

Getting regular feedback is key to maximising AI use. ‘Implementing feedback channels for clients and employees can provide valuable insights into policy effectiveness and support ongoing improvements,’ adds Wong. ‘This approach enhances financial integrity and strengthens client trust across their services.’

The path forward means walking a fine line between innovation and reliability through structured AI policies. ‘Developing AI policies not only addresses the AI governance issue, but is also a strategic priority that supports innovation and upholds a company’s reputation and professional integrity,’ says Yim.