A quiet revolution is under way inside the Tasmanian public service \u2013 and not everyone is fully on board.<\/p>\n
It began with the AI chatbot Microsoft Copilot, deployed since at least early 2024<\/a> and embedded in the pre-existing productivity software used by the state government.<\/p>\n Today, other platforms run alongside it, with still more being explored and trialled.<\/p>\n The Mercury contacted all government departments, government business enterprises (GBEs), and state-owned companies to ask them about their AI use.<\/p>\n In lieu of individual answers from the departments, a whole-of-government response was provided, which did not detail specific uses in each agency. <\/p>\n Nearly all GBEs and state-owned companies answered this masthead\u2019s questions.<\/p>\n Several government entities indicated they were now using additional AI tools to Copilot for administrative and analytical tasks.<\/p>\n A spokesman for TT-Line, the state-owned ferry operator, said the business had implemented its own AI policy and used Anthropic\u2019s Claude chatbot, plus the AI legal platform Plexus to manage contracts.<\/p>\n Other GBEs implied they were actively deploying, or exploring the potential to deploy, a range of AI systems but did not specify the exact tools to which they were referring.<\/p>\n The owner of Tasmania\u2019s mainline railways, TasRail, said it was in the early stages of investigating the use of AI for \u201clarge dataset modelling and analysis\u201d to help identify track geometry faults and improve fuel efficiency and conservation.<\/p>\n Innovation, Science and Digital Economy Minister Madeleine Ogilvie told the Mercury that Thursday\u2019s state budget would include $4m to \u201cidentify, test, and scale\u201d expanded AI use in the public service, with an eye to improving \u201cefficiency and productivity\u201d, while also reducing \u201cregulatory and administrative burden\u201d.<\/p>\n \u201cThe AI Accelerator project will work across agencies to drive adoption and implementation,\u201d she said.<\/p>\n Russell Kelly, the CEO of peak tech sector body TasICT, said AI was \u201cnot going to take people\u2019s jobs\u201d and workers who knew how to use it effectively would \u201cbecome more productive and more useful\u201d, provided they received appropriate training.<\/p>\n \u201cThe benchmarks we are seeing emerge are that AI can reduce the time taken to do some tasks by between 10-20 per cent,\u201d he said. <\/p>\n \u201cThis saving can be used on higher-value tasks that improve services for Tasmanians and the quality of decision-making.\u201d<\/p>\n In September 2024, the Department of Premier and Cabinet (DPAC) published guidelines for the use of AI in state government settings, which included transparency commitments and flagged the establishment of a whole-of-government \u201cAI initiative register\u201d. <\/p>\n Eighteen months later, this measure has still not been implemented, despite the public sector\u2019s increasing experimentation with the technology.<\/p>\n Mr Kelly said Tasmania\u2019s tech sector wanted to see AI implemented in the \u201cright\u201d way and he was \u201cconcerned\u201d the promised register had not yet been delivered.<\/p>\n Dr Alexandra Sinclair, a research fellow at the ARC Centre of Excellence for Automated Decision-Making and Society, said AI had the capacity to create remarkable societal benefits but the government needed to be upfront with the public about how exactly it was using these platforms<\/a>.<\/p>\n \u201cThe failure to institute \u2026 a register in Tasmania means there is no visibility within government of how AI is being used by departments and [it] does not allow for any public scrutiny or debate,\u201d she said.<\/p>\n Dr Sinclair said such a register should be made public and provide granular detail about how AI models were being used and the state should \u201cideally\u201d appoint an independent AI commissioner or oversight body.<\/p>\n \u201cThere is also a concern about the amount of power governments are giving private tech companies to facilitate government infrastructure and deliver government services,\u201d Dr Sinclair said.<\/p>\n \u201cThere should be transparency about this so the public can have a conversation around whether handing over power to those companies is desirable.\u201d<\/p>\n Ms Ogilvie said the AI initiative register had not originally been designed as \u201ca tool for risk management or compliance monitoring\u201d.<\/p>\n Since the publication of the state\u2019s AI guidelines, the function of the proposed register, due to be \u201cfinalised\u201d later this year, has been \u201creconsidered\u201d and would now be \u201cexpanded to better support governance and transparency in AI use\u201d, according to the minister.<\/p>\n A government spokeswoman said a \u201croad map\u201d and policy were being developed for the \u201cresponsible and effective\u201d use of AI in the public service.<\/p>\n Approved use of the technology within government departments currently involves secure, enterprise-grade software that is called on for such tasks as \u201cdrafting assistance, transcription, summarisation, research support, and workflow efficiency\u201d, the spokeswoman said.<\/p>\n Public servants are bound by privacy laws and are prohibited from entering personal information into AI platforms.<\/p>\n Community and Public Sector Union general secretary Thirza White said AI did not have to result in job cuts, as long as it was deployed appropriately and backed by \u201csignificant\u201d investment.<\/p>\n \u201cWhat we are seeking is an agreement between unions and the employer around AI that puts in place guardrails, guarantees no job losses, and establishes a training fund that ensures that any productivity benefits are shared with workers,\u201d she said.<\/p>\n \u201cWe are concerned … that government is encouraging workers to adopt these tools but not providing them with adequate training or risk frameworks – and then we\u2019re putting workers in a situation where they can inadvertently do the wrong thing.\u201d<\/p>\n The efficiency and productivity unit within DPAC is currently trialling an \u201cAI Assistant Platform\u201d using Microsoft\u2019s Azure OpenAI service, which features enterprise protections and boasts agentic capabilities that give users the ability to create autonomous agents that can automate decision-making and operate with limited to no human oversight.<\/p>\n The Mercury asked DPAC if agentic AI was being incorporated into the trial but did not receive a response by deadline.<\/p>\n