Dáil Éireann Debate, Tuesday - 21 April 2026
310. Deputy Malcolm Byrne asked the Tánaiste and Minister for Finance if he will consider the use of agentic AI in revenue collection by his Department and the State generally; and if he will make a statement on the matter. [26495/26]
Amharc ar fhreagraI am advised by Revenue that it makes extensive use of technology to support and improve operational efficiency. As the Deputy is aware, AI models are analytical tools that produce outputs such as summaries, forecasts, or draft text; they do not act unless a person decides to use their output. Agentic AI describes systems that use such models to plan and execute steps toward a goal, such as gathering information, calling approved software tools, or initiating routine processes, subject to permissions and oversight. Agentic AI is typically defined as AI which can perceive its environment, understand intention, and take action to achieve its goals. Agentic AI is not the same as AI agents nor autonomous AI.
I am further advised that Revenue has deployed AI models to assist in more accurately directing taxpayer queries to the appropriate Revenue caseworker, assist staff in accessing all relevant information to allow them promptly and accurately respond to taxpayer queries, to extract key data from documents and image-based material used by staff to expedite the 1st pass validation of information and generate initial drafts of Tax and Duty Manuals which are then rigorously verified and refined by subject matter experts before publication. The commonality across these models is that they are designed to support human decision making with a clear focus on gaining efficiencies. They assist professional judgment rather than replace it.
Revenue IT systems follow strictly defined business rules underpinned by the appropriate legislative and operational requirements. These systems are designed, developed, tested, scaled, and deployed to match those criteria. Agentic AI systems will likely have a role to play in these business processes and if appropriate will be built into business workflows and tested accordingly.
Finally, I would note that Revenue continues to monitor international developments in AI capability, including agentic systems, and engages with EU and OECD working groups on responsible AI deployment in tax administration. Any future applications involving greater system autonomy will be assessed through Revenue's established AI governance framework, including Data Protection Impact Assessments and alignment with the Government's Guidelines for the Use of AI in the Public Service.
