Increasingly, the first interaction citizens have with government services doesn’t involve a phone call or a front desk. It begins with a chatbot.
These digital assistants might answer licensing queries, route your complaint to the right department, or help you upload the correct document. In the best cases, they resolve queries in seconds. In the worst, they loop users back to the start with a vague prompt to “try again later.”
For regulators, this automation shift is not just about upgrading customer service, it’s about meeting the growing demand for scale, consistency, and speed. Governments are under pressure to serve more people with fewer resources. Citizens expect the same 24/7 responsiveness they get from their bank or insurer. And behind the scenes, regulators are juggling increasingly complex compliance obligations while managing large volumes of applications, complaints, and reviews.
The solution, for many, is automation – starting with chatbots and extending deep into the administrative machinery of regulation itself. But while the tools promise efficiency, their implementation raises fundamental questions about fairness, trust, and the limits of machine logic in human systems.
Public trust doesn’t depend on technical transparency so much as procedural reassurance. Most people don’t need to understand the inner workings of automation systems – but they do need to know that if something goes wrong, they can reach a real person who can help.
Chatbots in action: The new normal for engagement?
It’s one thing to talk about automation. It’s quite another to deploy it in the unpredictable terrain of public-facing regulatory work, where users may be stressed, multilingual, or digitally inexperienced, and where the right outcome often depends on subtle context.
Yet that hasn’t stopped a growing number of regulators from adopting chatbots as a first point of contact.
In Queensland, the government has developed an internal AI-powered assistant called QChat, based on OpenAI’s large language model. QChat is being rolled out across various departments to assist with tasks such as drafting briefing papers and refining emails. This initiative aims to reduce reliance on off-the-shelf generative AI solutions and address data security concerns.
Additionally, Queensland’s Department of Justice and Attorney-General introduced chatbots named MANDI and SANDI to assist residents with neighborhood disputes. These bots provide information on issues like noise complaints and fence disputes, allowing users to access services at their convenience.
The Australian Taxation Office (ATO) introduced a virtual assistant named ‘Alex’ to handle basic tax-related queries. In its first year, Alex managed over two million conversations, providing 24/7 assistance and alleviating pressure on human support channels.
In 2023, HM Revenue & Customs (HMRC) faced criticism for its customer service performance. The Public Accounts Committee reported that average call waiting times exceeded 23 minutes, with only 66.4% of customers’ attempts to speak to an adviser being answered, against a target of 85%. The committee expressed concern that HMRC had degraded its telephone service to drive taxpayers to digital channels, which could erode public trust in the tax system.
The Government Technology Agency of Singapore has developed the Virtual Intelligent Chat Assistant (VICA), a conversational AI platform designed to help government agencies respond to citizen inquiries efficiently. VICA leverages Hybrid AI, combining Natural Language Processing (NLP) and Generative AI (GenAI) to balance automation with accuracy. Currently, over 60 Singapore government agencies utilize VICA, hosting more than 100 chatbots that collectively handle an average of over 800,000 monthly queries. This widespread adoption underscores VICA’s role in enhancing public service delivery by providing 24/7 assistance and streamlining communication between citizens and government bodies.
Across the Pacific, the U.S. city of Denver launched an AI-powered chatbot named “Sunny” to assist residents with accessing city resources, reporting issues, and learning about local events. Available 24/7 on the city’s website, Sunny helps users schedule DMV appointments, report potholes or trash pickup problems, and inquire about various city services. While Sunny has improved access to information, the city acknowledges potential inaccuracies in the chatbot’s responses and disclaims responsibility for the generated content.
What all these examples share is a common logic: not every citizen interaction requires a person. If 80 percent of incoming questions are straightforward and repeatable, a well-trained digital assistant can absorb that workload, freeing human staff for complex, urgent, or sensitive work.
Inside the black box: Automating the back end
While chatbots dominate the public imagination, many of the most transformative automation tools in regulation are invisible. They live behind the scenes – streamlining workflows, routing complaints, validating forms, and guiding case officers.
In Australia, several regulators and compliance-focused organisations have adopted robotic process automation (RPA) to streamline data handling and reduce manual workloads. The Australian Prudential Regulation Authority (APRA) is replacing its legacy Direct to APRA (D2A) system with a new Data Collection Solution, part of a broader modernisation agenda to improve how it gathers and uses data from more than 4,500 financial institutions. This represents APRA’s most significant technology investment to date.
Australian Unity, meanwhile, implemented RPA to accelerate the processing of aged care assessment forms. The project reduced average transaction time from 30 to 4 minutes and reclaimed more than 22,000 hours of manual effort in under a year.
Decision engines are also gaining traction in regulatory systems. These tools apply policy logic and conditional rules – often structured as decision trees – to help determine outcomes such as whether a licence should be granted or a complaint escalated. New Zealand’s Ministry of Business, Innovation and Employment (MBIE) has implemented a low-code decision rules platform to streamline visa processing. The system enables policy staff to design and adjust eligibility pathways without coding, ensuring applications are triaged and assessed against up-to-date policy logic.
None of this removes the need for human judgment. But it does reshape where and how that judgment is applied.
Risks and cautionary tales
For all their promise, automation tools in regulation carry risks. When deployed carelessly or without accountability, they can frustrate users or erode public trust.
Some early chatbot rollouts gave the impression they could resolve anything. In reality, they couldn’t. In HMRC’s example, efforts to shift users toward digital channels drew criticism as customer service performance declined and wait times ballooned. While not entirely due to automation, the shift highlighted the risks of deprioritising human support too aggressively.
In Australia, rules-based automation has delivered efficiencies in areas like data collection and licensing, but it has also raised concerns – such as in Queensland’s health sector, where a triage system used by the Office of the Health Ombudsman was criticised for prematurely closing complaints without referral to national regulators.
Bias is another concern. When risk models rely on historical data, they can unintentionally reinforce existing enforcement inequities – especially if past decisions reflect systemic imbalances. And performance metrics tied to automated workflow speed can unintentionally pressure staff into hasty case closure.
That’s why successful regulators proceed with caution. They start small, involve staff in tool design, and offer escalation paths.
Where humans still matter – and always will
What automation shouldn’t signal to regulators is the notion people will no longer be a part of the equation. Instead, it’s about supporting them. Optimal systems leave room for human judgment.
New Zealand’s government AI framework encourages the use of automation to support, not replace, human decision-making – a principle that underpins digital services across agencies including MSD and Inland Revenue.
Internally, automation changes how people work. It brings consistency – but also rigidity. Change management is essential. Queensland’s Digital Services Standard requires agencies to co-design digital tools with frontline staff.
Ethically, regulators must ensure transparency, contestability, and fairness. That means clear communication, accessible appeals, and assurance that no one is left behind.
What’s next? Smarter tools, better design
Looking ahead, automation will become more context-aware. Chatbots will link queries, decision engines will use richer data, and workflow tools will surface insights.
Multilingual bots are expanding, particularly in New Zealand. Predictive analytics are emerging, helping regulators detect early compliance risks. Tools are beginning to shift from reactive support to proactive oversight – identifying patterns and triggering interventions before issues escalate.
Australia’s Digital Transformation Agency (DTA) and the New Zealand Government Digital Services teams are both exploring frameworks for responsible automation. These include privacy-by-design principles, regular audits, and transparency obligations when AI or automation is involved in decision-making.
But fundamentals remain unchanged. Good automation solves real problems with clear outcomes and human oversight. It’s well-governed, tested rigorously, and built with the end user – both public and staff – in mind.
Automation may transform how regulation is delivered – but it’s human judgment, trust, and accountability that give regulation its purpose.
