The tAIxman: how AI is transforming Ireland’s revenue collection
Finance departments can be slow to adopt new technologies but in the land where the computer was first conceived, Ireland’s Revenue agency is finding a host of applications for artificial intelligence – from customer service to records management, and from producing manuals to testing legislation. Matt Ross reports
“Ireland has a really long connection with artificial intelligence,” said Ruth Kennedy, a revenue commissioner in Ireland’s tax and customs agency. When Jonathan Swift settled in County Tyrone to write the 1726 novel Gulliver’s Travels, she recalled, he included a mechanical sentence-generator named ‘The Engine’ – becoming the first person to describe a computer. In 1956, Kennedy continued, maths professor John McCarthy – the grandson of an Irish immigrant from County Kerry – organised the Dartmouth Summer Research Project on Artificial Intelligence (AI) in New Hampshire, arguing that “every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it”.
Nearly 70 years after his namesake invented the very phrase ‘artificial intelligence’, a contemporary John McCarthy was present as those attending the Global Government Finance Summit debated ‘how finance ministries can use artificial intelligence’. During the previous evening, the chief economist at Ireland’s Department of Finance welcomed senior finance leaders from 12 countries and the European Commission to Dublin for the Summit. In a speech covering the structural changes affecting economies around the world, McCarthy highlighted how AI is extending automation into new swathes of the working world. “We used to have machines replacing hands,” he commented. “We now, potentially, have machines replacing heads.”
National treasuries and economics departments must take the lead in trying to understand how AI will affect investment, productivity, employment and tax revenues – but one civil service leader suggested that many finance functions are proving slow to implement AI technologies themselves. “According to a study by Gartner, finance functions lag behind in the incorporation of AI,” they said. “Thirty percent of them have no planned AI implementation, and only 8% are currently using AI in production.”
An adaptable assistant
This is odd, given AI’s capabilities in managing numerical data: the technology has obvious applications in fields such as budget forecasting, automated data extraction, predictive analytics, fraud detection and customer service. As Kennedy had – a little wryly – argued, Ireland has long been on the front foot on AI. And in the modern age, she argued, Ireland’s Revenue agency has maintained that tradition. “As a tax and customs administration, we’re a very data-rich organisation. We moved online back in 2000, when we introduced the ability to file and pay all your taxes and import and export declarations online,” she said. “So we have a wide source of data available, and we’ve used artificial intelligence for many years.”
Revenue has found particular value applying AI in customer services, Kennedy explained. The agency used to provide citizens with a drop-down menu to categorise and direct their queries, for example, but now provides a free-text box and uses AI to decide which department should answer questions. The proportion of queries landing with the correct department has since jumped from 70 to 98%. Revenue has also launched a Large Language Model (LLM) able to feed information to the staff answering citizens’ queries online and on the phone. Trained on Revenue’s curated information – and thus avoiding the risk of ‘hallucination’ – this “Revenue helper” is now available to all 7,000 agency staff. The system links back to all its source material, Kennedy explained, “giving people the ability to fact-check, use their judgement and make sure the answer is correct”.
Read more: The ‘four Ds’: Ireland’s chief economist spells out the major challenges governments face
More AI applications
Revenue is also using AI to summarise call centre conversations, saving staff that time-consuming task. The system records and transcribes the call, then produces a brief summary for approval by the call handler. Further efficiencies will be generated by combining these systems, Kennedy continued: callers will soon be asked to describe their problem while they wait for someone to answer the phone – by which time the call handler will have been briefed on the caller’s questions and given suggested answers.
In a further application of AI, the agency is working on how an LLM could “create the first draft of the tax and duty manuals” produced to explain new laws to citizens and businesses, “using as its inputs the legislation and the notes provided to the minister.”
It’s already used similar models to pull out the common themes in responses to consultation exercises, and to draft summaries of long, complex documents such as High Court judgements.
Predictive analytics has long proved a fruitful area for Revenue, enabling staff to identify anomalies in large datasets that might suggest incidents of fraud or error. More recently, the agency has begun training AIs to look for weaknesses in draft legislation: “Can we try and break the legislation? What loopholes, what avoidance, what evasion could be found in there?”, Kennedy asked. Such exercises will become ever more important, she added, noting that tax advisers also have access to such tools. “If you look at the large law firms, they’re all launching AI in their suite of applications to review both judgements and legislation,” she said. “Even in a room of 10 individuals trying to scrutinise something, there are things that may not come up; AI potentially can surface some of them.”
Reorganising around the new recruits
So does all this new technology augment or replace civil service staff, asked John McCarthy. “And if you need fewer of them, is it freeing up those workers to do more cognitive or more technical or more highly skilled jobs?” Right now, the AI systems are “supplementary, because we’re still very much in the proof-of-concept stage”, replied Kennedy. “But I can already see the exponential change it will mean. Ultimately, we’d like 98% of queries to be dealt with by AI.”
Some citizens will always require a human interaction, she added, but many of the queries handled by Revenue could easily be addressed by existing AI technologies. Over time, she anticipates a reduction in the size of the workforce – “but that doesn’t mean we’ll be cheaper because things are becoming more complex”, Kennedy added. “We’ll need smaller numbers of higher-paid staff in future.”
As these technologies spread throughout the operation of finance departments, staff will need a new set of skills. In particular, Summit participants highlighted the importance of training senior leaders in the nature and capabilities of AI technologies – building the confidence and skills required to design, commission and manage these new tools. Helpfully, commented Kennedy, LLMs can help to create training materials – even producing “targeted training, based on the way you like to learn and your level of knowledge – so very much personalised”.
Lessons learned
Other finance departments interested in pursuing AI, said Kennedy, should bear a few key principles in mind. First, “identify the pain points”: places of “limited resources and competing demands, where people are saying: ‘I can’t answer all these phone calls!’” Pull in ideas from the whole workforce: “We’ve been doing roadshows all over the organisation to get people thinking about it,” she said. Then consider the costs and benefits of intervening, feeding the results into a strong governance and prioritisation process. Define and measure the change you’re seeking and remember that, unlike traditional ICT projects, AI systems generally lean heavily on revenue rather than capital spending – requiring ongoing payments for processing capacity, data handling and algorithm management.
In developing projects, Kennedy continued, make sure you involve all those with a stake in the system: “Bring operational people, legislative people, technology people into multidisciplinary teams, and run collaborative sessions to understand the potential use cases and get buy-in from all sides,” she said.
Finally, when building the new systems, hold projects close enough to retain control and ownership of key assets. “As a civil service organisation, you can’t get all the ICT skills you need at the price point we’re able to pay,” she said. “So we’ve always supplemented that with external resources from some of the big tech consulting firms, but we’ve been very conscious to keep the architecture, the data and the governance in the hands of civil servants.”
Retaining trust
How is Revenue adapting to the EU’s new AI Act, asked Declan Costello, deputy director-general of the European Commission’s Economic and Financial Affairs Directorate. “Has the right balance been struck between addressing the key public policy concerns, while also allowing innovation and efficiency?”
Revenue is developing governance structures to comply with the Act, replied Kennedy, adding that the Act is an important tool in building the public trust required to realise the potential of AI. “Our mission is to serve the community. We want to achieve that by voluntary compliance – and how we do that is by being trusted,” she said. “So reputation is really important for us.” The AI Act, she said, can help secure the public’s buy-in as these new tools appear in public services.
“I think Europe is leading the way on this, and bringing the rest of the world with us. Because so many European organisations and companies will use the Act, it will by default ripple out into how services are provided,” she added. “I always say that with great data comes great responsibility, and it’s critical for us that we keep the trust of the community and of taxpayers.”
The invitation-only Global Government Finance Summit is a private event, providing a safe space at which the civil servants leading finance departments and national treasuries can debate the challenges they face in common. We publish these reports to share some of their thinking with our readers: note that, to ensure that participants feel able to speak freely, we give all those quoted the right to review their comments before publication.
The 2024 Meeting will be covered in five reports, covering the various sessions:
– John McCarthy, Chief Economist of Ireland’s Department of Finance, analyses the structural challenges facing advanced economies
– How finance ministries can use artificial intelligence
– Making tax less taxing: using technology to improve assessment and collection
– How to improve public sector productivity
– Using ‘just in time’ insight in government finance and service delivery
For information on the next Global Government Finance Summit, which will be held in June 2025, visit our dedicated website.
About Matt Ross
Matt is Global Government Forum's Contributing Editor, providing direction and support on topics, products and audience interests across GGF’s editorial, events and research operations. He has been a journalist and editor since 1995, beginning in motoring and travel journalism – and combining the two in a 30-month, 30-country 4x4 expedition funded by magazine photo-journalism. Between 2002 and 2008 he was Features Editor of Haymarket news magazine Regeneration & Renewal, covering urban regeneration, economic growth and community development; and from 2008 to 2014 he was the Editor of UK magazine and website Civil Service World, then Editorial Director for Public Sector – both at political publishing house Dods. He has also worked as Director of Communications at think tank the Institute for Government.
Related Posts
