Hong Kong is moving artificial intelligence into one of its most visible public-service channels. The city’s Digital Policy Office has announced four planned AI projects for 1823, the government platform that handles public inquiries and complaints. South China Morning Post reported that the planned system has raised questions about data safeguards even as officials expect it to reduce frustration for users. The government’s own release confirms that the work will cover call transcription, email extraction, prioritization of departmental replies, and a dashboard for local issues.
The plan matters because 1823 is not a small trial service. The Digital Policy Office says the platform handled more than 7.5 million inquiries and complaints last year. That scale turns even a modest reduction in administrative work into a significant operational change. The office expects to begin rolling out the new tools this year, and its planned AI-agent pilot inside government is scheduled for the fourth quarter.
Hong Kong already has a broad policy framework for using artificial intelligence in government and financial services. The city’s new 1823 projects add a day-to-day public-service layer to that agenda. They also raise a practical question that cannot be answered only by a product demonstration: whether a government can use language models to speed up routine work without weakening accountability, confidentiality, or the right of residents to have difficult cases handled by people.
Four AI Projects Target Calls, Emails, Case Priority, and Local Issues
The Digital Policy Office named the four projects as Call Assistant, Email Extraction, Prioritisation of Departments’ Replies, and Dashboard. Each is tied to a particular bottleneck in a large contact center. Call Assistant will use speech recognition and large language models to transcribe calls on specific topics in real time and generate case summaries. The office estimates that this could reduce after-call work by 15%.
Email Extraction is intended to analyze incoming public emails and automatically pull case data into the customer-service management system. The office projects a 30% reduction in manual data-entry time. That is a narrow but useful application of language models: instead of asking a model to decide a complaint, the system is being used first to turn unstructured correspondence into fields that staff can review and route.
The third project focuses on the time lost before a department addresses a potentially urgent issue. The Digital Policy Office says AI will analyze emails and identify cases requiring priority handling, reducing identification time from two days to a few hours. The fourth, Dashboard, will combine information from livelihood-related cases and display it visually, referring issues to responsible district-level departments for follow-up.
These are administrative functions rather than a decision to automate public judgment. The line is important. A summary, a data-extraction tool, or an alert can help a civil servant manage a large queue. It should not become an unreviewed answer to a resident whose case involves housing, welfare, public safety, or a dispute with a government department. The government’s description places the projects inside 1823’s existing case-management process, and that structure should remain visible as the tools are introduced.
Hong Kong has already been confronting the public-trust dimension of AI. EastFrontier previously documented a rise in privacy complaints in Hong Kong’s AI era, a reminder that faster digital services can also increase concern about where personal information is stored and how it is used. The 1823 plan will be judged not only by time saved but also by whether the office explains its data controls and keeps meaningful human review in place.
The 1823 Rollout Links Service Automation to Government AI Infrastructure
The 1823 tools are part of a wider Digital Policy Office program to support AI adoption across Hong Kong departments. According to the official announcement, the Government Cloud Infrastructure Services provide private-cloud capacity, shared AI computing resources, large-language-model tools, and development tools for text understanding, image generation, knowledge-base creation, and process automation. The services support more than 60 bureaus and departments and power more than 600 digital government services.
Since July, those private-cloud services have also supported document-assistance tools developed by the Hong Kong Generative AI Research and Development Center. The office says those tools help with drafting documents, translation, meeting minutes, and smart searches, including work involving confidential materials. That context is relevant to 1823 because it suggests the new projects will not be built as isolated consumer applications. They are intended to operate within government infrastructure and governance rules.
The next public-facing step is also scheduled for the fourth quarter. The Digital Policy Office plans to launch an 1823 web application on the iAM Smart platform, replacing the existing 1823 mobile app. It also plans an iAM Smart AI Assistant that will let residents ask questions in everyday language and receive information and service links from the mobile platform. Separately, government AI Agents will be piloted for standardized and repetitive administrative tasks.
Those plans place Hong Kong in a different position from a conventional chatbot deployment. The government is connecting a contact center, identity-linked service platform, private cloud, and internal workflows. EastFrontier’s recent coverage of Hong Kong’s 36-project financial AI sandbox showed that the city is also using controlled environments to test AI in banking, securities, and insurance. The 1823 work extends the same cautious deployment logic into citizen services.
Data Safeguards Will Determine Whether Residents Trust the New System
SCMP’s reporting highlighted expert concerns that data safeguards will be essential as 1823 adopts AI. That concern follows directly from the kinds of data the system handles. Public complaints can contain phone numbers, addresses, personal histories, health information, employment details, or allegations about other people. A system that transcribes calls and extracts case data is useful precisely because it can process such information at volume.
The Digital Policy Office has emphasized a “prudent and progressive” approach to AI adoption. In practice, that should mean clear rules on retention, access, model training, vendor controls, and the escalation of uncertain outputs to staff. It should also mean avoiding an assumption that a tool that works well on routine emails will work equally well on multilingual calls, emotional complaints, or cases that require legal interpretation.
The program has measurable targets, including 15% less after-call work, 30% less manual entry, and a reduction in urgent-case identification from two days to hours. Those targets create a useful basis for later scrutiny. Residents and legislators will be able to ask not only whether the system saved time, but whether it improved routing accuracy, reduced repeat contacts, protected data, and gave officials a traceable record of how a priority label was assigned.
Hong Kong’s 1823 AI plan is therefore best understood as an operational test of public-sector AI rather than a simple chatbot launch. It applies large language models to a service with 7.5 million yearly interactions, connects them to public cloud infrastructure and iAM Smart, and makes efficiency claims that can be measured. If the city can deliver the promised gains while retaining privacy protections and human accountability, the project could offer a practical model for AI-assisted government services. If not, the same scale that makes it attractive will make its failures highly visible.