By Tony Low · AI in Action · October 2, 2026

If you’ve ever called a government hotline and wondered whether anyone would actually read your complaint, Hong Kong’s 1823 service is worth a look. It’s the city’s single phone number and portal for everything from noisy air-conditioners to blocked drains, and last year alone it handled 7.55 million enquiries and complaints. On 27 August, the city’s Digital Policy Office announced four new AI tools designed to make that firehose of citizen contact easier to triage, route, and resolve — and the infrastructure choices behind them are as interesting as the tools themselves.

The four projects, called AI Efficacy Enhancement Projects, each target a distinct bottleneck in the hotline’s workflow. A Call Assistant uses speech recognition paired with a large language model to transcribe calls in real time and auto-generate case summaries, which the Office estimates will cut after-call administrative work by 15%. An Email Extraction tool uses AI to pull structured case data out of incoming emails and feed it directly into the case management system, targeting a 30% reduction in manual data entry. A Prioritisation tool flags which departmental replies need urgent attention — collapsing a process that used to take two days down to a matter of hours. And a consolidated Dashboard pulls livelihood-related cases into one view so district-level departments can act on them faster.

None of this is framed as a moonshot. It’s unglamorous, plumbing-level automation — exactly the kind of AI deployment that tends to get less attention than flashy chatbot launches, but that actually changes how civil servants spend their day. Acting Commissioner for Digital Policy Daniel Cheung put it plainly: “AI is rapidly transforming the mode of public services,” he said, while stressing the Office’s intent to pursue adoption safely and methodically rather than all at once.

What makes this more than a routine software upgrade is what it’s built on. The four tools run on the Government Cloud Infrastructure Services, the same backbone that already supports more than 60 government departments and over 600 digital services — kept under the administration’s own control rather than handed to an outside vendor’s stack. And the document-assistance components weren’t bought off the shelf: they were built with the Hong Kong Generative AI Research and Development Center, a locally-based R&D outfit whose tools have reportedly been running inside government workflows since July. That combination — public-sector AI adoption delivered on government-controlled infrastructure, co-developed with a homegrown research centre — is precisely the kind of quieter sovereignty story that doesn’t always make headlines next to national LLM launches, but matters just as much for a government handling millions of residents’ data.

The rollout isn’t finished. The Office has pencilled in Q4 2026 for two further pieces: folding the new AI capabilities into the iAM Smart mobile app, so residents can query government services in natural language instead of hunting through menus, and piloting internal “AI Agents” that can act on staff’s behalf inside departmental systems rather than just summarising information for them. That’s a meaningful jump — from AI that assists a human doing the work to AI that does a constrained slice of the work itself — and it’s the step worth watching most closely into next year.

It’s a useful case study for anyone building public-sector AI elsewhere in the region: the headline-grabbing part of a sovereignty story is usually the model, but the durability of this one is in the boring parts — whose cloud it runs on, who wrote the fine-tuning pipeline, and whether a local R&D centre, not a foreign vendor, is the one iterating on it next quarter. Hong Kong’s 1823 overhaul is a reminder that a lot of the most consequential AI deployment in APAC right now isn’t a press-conference LLM — it’s the unglamorous automation sitting quietly inside a contact centre, handling millions of calls a year a little bit faster than it did last month.

Source: Government of Hong Kong SAR press release, 27 August 2026; OpenGov Asia, August 2026; The Standard, June 2025.

Tony Low covers how AI gets built, deployed, and used across the region for AIAPAC.ai Insights.