Hong Kong is expanding the use of artificial intelligence on its 1823 inquiry and complaint platform in an effort to reduce waiting times, improve the handling of cases and strengthen governance, but the move is also raising questions about privacy, data security and the risks of allowing increasingly autonomous systems to make administrative decisions.
The Digital Policy Office, which operates the platform, said it would expand AI-assisted functions to transcribe residents’ calls, summarise emails and record information in its system. The technology will also be used to triage complex and urgent cases for experienced staff to forward to relevant departments, while identifying and summarising emerging district governance issues from residents’ complaints.
The objective is to allow staff to spend more time dealing with complicated cases and responding to residents’ needs. Technology-sector lawmaker Duncan Chiu said the expanded capabilities could address a persistent source of public frustration: complaints being sent to the wrong authorities. He pointed to cases involving issues such as fallen trees on boundaries between private and public land, where responsibility can become unclear between departments.
Chiu said AI could help operators ask more targeted questions during calls, gather clearer information and direct complaints more accurately from the outset. Such a system, he argued, could reduce unnecessary transfers between departments and save residents significant time.
The delayed adoption of AI on the platform, however, reflects technical challenges specific to Hong Kong. Chiu said few large language models had been able to reliably transcribe Cantonese, particularly because of the language’s range of accents and the frequent mixing of Cantonese with English or other languages. Even seemingly simple references to local landmarks can require additional training for AI systems to understand accurately.
Chiu was among 67 lawmakers who visited Beijing during the summer and toured the 12345 Citizen Hotline Service Centre, where big data and AI are used to route complaints dynamically to government authorities, with deadlines established for resolving cases. He said Hong Kong’s technological initiatives could match about 80 per cent of Beijing’s platform capabilities, but argued that the major differences extended beyond technology.
Beijing’s system, Chiu said, benefits from top-down administrative mandates requiring departments to resolve assigned complaints within strict 14-day deadlines, together with a fully integrated and centralised cross-departmental database. Hong Kong authorities may therefore need to determine whether adopting a similar approach would be cost-effective.
Lawmaker Johnny Ng Kit-chong said the expanded use of AI was a useful starting point and suggested that the technology could eventually play a wider role in monitoring how departments respond to complaints. AI could also analyse photographs and videos submitted by residents using computer vision, while automated systems could follow up with departments until action was taken and the complainant was satisfied with the outcome.
Michael Chau, a professor of innovation and information management at the University of Hong Kong, said combining AI with human involvement could offer a better balance between speed and quality than relying exclusively on either. While the effectiveness of the upgraded system remained to be seen, he said AI could significantly shorten the time residents waited for responses.
The more consequential experiment will come in the fourth quarter of this year, when the Digital Policy Office plans to pilot agentic AI. Unlike conventional AI tools, agentic systems can act autonomously on users’ behalf to carry out tasks, potentially giving them a greater role in administrative workflows.
Chau said the trial reflected a wider industry trend towards using AI agents to automate workflows and allow workers to focus on higher-level tasks. But he warned that greater autonomy also creates greater risks. Because human supervision is reduced, government systems handling sensitive information would require safeguards against data breaches and unintended actions by AI agents.
Chiu supported the trial, calling it a logical and timely response to the “explosive” growth of the technology. He said advances over the past year had addressed some of the earlier concerns surrounding data privacy and security and expressed confidence that the government could use AI agents safely to transform administrative workflows.
For Hong Kong, the expansion of AI on 1823 therefore presents a dual challenge. The technology promises faster and more accurate complaint handling, but its success will depend not only on whether machines can process cases efficiently, but also on whether authorities can maintain effective human oversight and protect the sensitive information entrusted to the system.
