Hong Kong’s 1823 AI Upgrade Signals a Wider AIO Shift

Seven and a half million. That is how many enquiries and complaints the Hong Kong government’s 1823 service handled last year, and from the fourth quarter of 2026 a growing share of the work behind them will be done by AI.

The Digital Policy Office set out the plan on 27 August 2026, announcing four AI Efficacy Enhancement Projects for the service alongside broader support for AI adoption across bureaux and departments. Read as a technology story it is modest. Read as a signal about what residents will soon expect from any organisation answering a question, it argues strongly for taking AI optimisation (AIO) seriously. The acronym is used in two ways in this industry, so to be clear: AIO here means optimising your information so AI systems can find, understand, and reuse it, not Google’s AI Overviews.

What the Government Is Actually Building

The four projects are narrow and measurable, which is what makes them credible.

Project	What it does	Stated target
Call Assistant	Supports operators during and after calls	Reduce after-call work time by 15 per cent
Email Extraction	Pulls structured details out of incoming email	Reduce manual data entry time by 30 per cent
Prioritisation of Departments’ Replies	Identifies which replies need attention first	Cut identification time from two days to a few hours
Dashboard	Consolidates and visualises livelihood-related case data	Not stated

 

Each project targets clerical load instead of judgement, which is the sensible place to start and the reason the numbers are believable. Nothing here replaces an officer’s decision about a complaint.

Acting Commissioner for Digital Policy Daniel Cheung framed the wider ambition in the announcement: “AI is rapidly transforming the mode of public services.” The same release noted that Government Cloud Infrastructure Services now supports over 60 bureaux and departments, powering more than 600 digital government services, which gives the AI work a platform to spread across.

The Expectation Reset

Public services set the floor for what people consider normal. A resident who gets a fast, accurate, AI-assisted answer from a government line at nine in the evening does not then accept a two-day email turnaround from a private company, and the South China Morning Post reported the vetting expansion arriving by the end of the year. That is the part Hong Kong businesses should sit with, because the competitive baseline for answering a question is being reset by an organisation nobody thinks of as a competitor.

What AIO Asks of a Business

AI optimisation is less exotic than it sounds. An AI system, whether it belongs to a search engine or to a company’s own support stack, can only work with information it can locate and parse. Most Hong Kong businesses fail at the locating step long before the parsing one.

The practical work tends to look like this: publish your service details, pricing logic, coverage areas, and turnaround times as text on your own site instead of burying them in a PDF or an image. Keep one canonical version of each fact, so an agent reading your site does not find three different opening times. Write answers to the questions customers actually ask, in the words they use, and keep the answer at the top of the page.

None of that requires a platform decision or a new vendor. It requires someone to own the accuracy of the company’s own information, which is a job that rarely appears on an organisation chart. Two habits separate the businesses that get this right. They keep a single owner for every published fact, so corrections propagate instead of accumulating, and they write for the question rather than the search term, which produces content a model can lift a sentence out of cleanly.

The signals in IAB Hong Kong’s AI Marketing Playbook pointed the same way earlier this year, well before the government put figures against it.

How the Corridor Diverges Here

Singapore has been running at this from a different angle, and the contrast is instructive for anyone operating in both markets. Singapore’s public sector has focused on agentic experiments and enterprise capability building, with the Infocomm Media Development Authority documenting AI agent work with the Singapore Government and a Google sandbox. Hong Kong’s announcement is squarely about throughput in a service that already exists at scale.

Neither approach is the better one, and a brand expanding in either direction should notice the difference in what gets rewarded. In Singapore, an English-first market with heavy Google reliance, information published cleanly on your own site does most of the work. In Hong Kong, the same content has to hold up in Traditional Chinese as well, and once a business moves across the boundary into the mainland it faces platforms with entirely separate rules about what they will read and cite.

The corridor lesson is that AIO is one discipline applied against three different reading systems. A Singaporean company treating Hong Kong as a copy-paste market usually discovers this at the point where its Chinese-language pages turn out to be an afterthought.

Where to Begin

Ask someone to search for your business the way a customer would, then check whether the answer they get is current. That single exercise finds more problems than most audits, and it costs an afternoon.

If you would like a hand turning your own information into something AI systems can read reliably, across English and Traditional Chinese, our team is based in Singapore and works across these markets daily. Get in touch and we can start with your search foundations.