Tax Technology Conference 2025: the current and future landscape

More than 250 tax professionals attended the ATT-CIOT Tax Technology Conference 2025, held at Birmingham’s International Convention Centre on 4 June. Conference sessions explored the current and future tax technology landscape, and what it means for tax practices in terms of preparation, safe adoption and advancing ethical AI in tax.

The day was top and tailed by the two organisations’ chief executives. CIOT’s Helen Whiteman, who conceived the idea of the conference, welcomed attendees with some of the findings of a survey of their ‘AI level’. This ranged from the discovery and learning phase (roughly half) through to a small number fully using AI tools already. ATT’s Jane Ashton wrapped up the day, thanking speakers and attendees and setting out some of the ways the two organisations are adapting their offer to students and members to reflect technological advances.

In between, attendees heard from leading figures in the adoption of AI and other new technology by HMRC and tax practices, took part in interactive workshops and were shown demos of cutting-edge products. The day closed with a drinks and networking reception.

Keynote Address: Professor Michael Mainelli

In his keynote address, former Lord Mayor of London, Professor Michael Mainelli, told attendees that AI will make tax more efficient by automating tasks, improving processing speed and reducing costs, and more effective by enhancing compliance, targeting fraud and informing better tax policies. But he warned them to exercise caution and understand the flaws of AI, which he described as ‘like a very happy puppy’ that is so keen to please you that it will serve up whatever it thinks you want – even if this sometimes means making things up.

Professor Mainelli suggested that to fully attain the benefits of AI, it is first necessary to simplify the tax system. ‘Years ago, when leading a research team in the automotive industry, I learned a mantra about computerisation – simplify, then automate, then integrate… AI can take huge balls of information knitting wool and tangle it better than any kitten. If we want to reap the benefits of AI, we should truly look at simplifying our tax code and procedures before we automate and integrate them.’

Later in his speech, he warned that Central Bank Digital Currencies (CBDCs) are ‘potentially a game changer’ when it comes to tax collection powers, giving policy-makers new tools to complicate the tax system. ‘Complex taxation algorithms can be applied to any CBDC transaction in real time… Once people realise the power of CBDC systems to support various taxation initiatives at low transaction costs, we could expect avalanches of new taxes [to be] proposed… You could easily have child noise taxes, alcohol consumption taxes, foreign visitor taxes, plastic bottle taxes, and so on.’ He said he had once given an example of such a tax to a parliamentary committee: the Nelson’s Column tax, a hypothetical populist redistribution tax on transactions, set eye-wateringly high in central London and declining as you head out of the capital. CBDCs would make such a tax possible, he suggested.

Professor Mainelli also warned about what he called the ‘illusion of innovation’. 25 years ago, automated legal discovery tools were supposed to do lawyers out of work, but instead just increased the scale of documentation enormously. ‘Now we talk about AI doing lawyers out of work… [But] AI is now helping them prepare ever longer briefs of dubious quality, and review ever longer briefs from the opposing side.’ Innovation, he concluded, does not always result in benefits.

Technical Tax Updates

The morning panel session saw Craig Ogilvie, HMRC’s Director of Making Tax Digital (MTD), tell the conference how HMRC are building on the success of their app and how we need to ‘watch this space’ for the digital transformation roadmap. He emphasised that quarterly reporting was being made as simple and cumulative as possible. In response to questions, Ogilvie assured attendees that there are no plans to require taxpayers to pay income tax quarterly. Similarly, there is no immediate plan to introduce MTD for partnerships.

ATT’s Director of Public Policy, Emma Rawson, noted that while MTD is the headline, there are also lots of other technological advances coming in tax administration. One of these is e-invoicing, where Rawson stressed the need to weigh the benefits against the burdens, arguing that the government should push a voluntary rollout rather than mandation. She was unconvinced by the idea of triangular e-invoicing via the tax authority, suggesting that where it had been tried it was usually countries with a big tax gap to close, where the burden it imposes might be seen as proportionate. That was not the case in the UK, she said.

Jane Mellor, Head of Professional Standards for CIOT and ATT, spoke about the implications of AI for good practice and professional standards – for example, the need to check output before sending it on and the importance of not letting client data end up in the data sets of tools like ChatGPT. Similarly, when an adviser gives advice that might be put through AI, what are the issues there? The professional standards team are currently working on topical guidance on the application of Professional Conduct in Relation to Taxation (PCRT) principles in relation to use of AI.

Mellor also reminded us that when self-assessment came in, some people said it would mean fewer tax advisers, which did not seem to have transpired, so we should be wary of suggestions that that will be the case with AI.

Breakout Sessions

Attendees took part in four breakout sessions during the afternoon.

Anna Kwiatkowska and Nicola Smith of HMRC presented on AI at HMRC. They identified five ways in which HMRC uses AI today:

• assessing the risk of non-compliance;
• the Ask HMRC Online chatbot;
• analysing customer feedback and contact data;
• recommending debt recovery actions; and
• directing correspondence.

They explained that HMRC’s IT strategy envisages a much simpler future technology landscape, with fewer separate applications and buying more of the IT the department uses as a service from suppliers.

Participants formed groups to provide feedback to HMRC. Recurring themes included the need for agents to be able to see and do everything the taxpayer can, for a way to track correspondence for agents and for a secure messaging service within the agent services account. Issues caused by the lack of interaction between different HMRC systems were identified.

Dr Sam De Silva of CMS and Matt Woolgar of PwC presented on Integrating AI solutions in professional services firms. This covered how to adopt AI tools within firms, including deciding which tool best fits a firm’s needs. Attendees were asked to think about what they do as a business, and what services they provide clients, when deciding which tools to incorporate. This included general productivity, summarisation, technical research, data extraction and analytics. The session also focused on legal and contractual issues around adopting AI tools, including who takes on liability if issues arise.

Xiaoshan Sun, Tax Technology Lead at Deliveroo, and John Sandall, CEO of Coefficient, hosted a session on Best practices for implementing tax technology. The session highlighted current AI tools and what they can do for small business owners and practitioners, including key prompting techniques and a discussion on what can go wrong when using AI. Attendees were walked through the AI Management Essentials Tool, which is being developed to help small and medium enterprises to establish robust management practices for the responsible development of AI systems and use of AI products.

Rob De La Rue of RSM UK Tax and Russell Gammon of Tax Systems brought together their wealth of experience to present A practical guide for using technology within in-house tax teams. The session highlighted that access to technology tools may already be available to businesses via their existing software, as well as demonstrating bespoke software. The presenters highlighted that the journey to embedding tax technology within a business is normally driven by obligations for VAT and corporate tax reporting, and this brings opportunities to deliver greater analysis and controls for the tax team, as well as producing accessible insights about tax in the business for non-tax colleagues.

Panel Discussion: Ethics and AI in Tax

The afternoon panel saw a discussion on the ethics of AI, chaired by CIOT Deputy President Paul Aplin.

Graham Tilbury, a Partner at the advisory firm WTS Hansuke, raised the problem of AI bias, stressing the need for transparency. Anna Kwiatkowska, HMRC’s Chief Data Scientist, acknowledged this issue but pointed out that humans too have biases, and at times AI models are less biased than people. The key lies in the training and skills of users, she suggested.

Dr Sam De Silva, a Partner at the law firm CMS, talked about the importance of ‘explainability’, being able to explain decisions made on the basis of objective data. He observed that if you apply for credit and are turned down, if the decision-making has been fully automated you have a right under GDPR rules to ask for a human to review that decision. Kwiatkowska sought to reassure practitioners that wherever AI results in outcomes for HMRC’s customers there is ‘always a human in the loop’.

Aplin highlighted the risks of AI hallucinations, noting that while large language models often get things impressively right, they sometimes get them badly wrong. Observing that HMRC is now using AI to deal with taxpayer queries, Aplin asked who is responsible if an answer is wrong?

De Silva said that if a taxpayer acts on wrong information they receive from HMRC’s chatbot, the tax authority should be responsible for that, but he expected that they would disclaim liability. Priya Vijayasarathy, Director of Data and AI at Deloitte, said that the end goal of a correct tax return hasn’t changed so the onus to get it right is always going to remain on the taxpayer. Kwiatkowska agreed. She explained that HMRC’s own chatbot produces answers based on curated content, operating within very strong guardrails. GOV.UK is trialling a generative AI chatbot but feedback suggests it’s getting some answers wrong.

Vijayasarathy stressed the importance of putting good governance in place whenever you replace human tasks with AI. De Silva posed the question of whether people would trust that information they put into HMRC’s chatbot – perhaps researching the implications of something they are considering – won’t be used by HMRC. Aplin suggested that, in order to know that AI is answering the right question, the default should be to play back not just the question but the assumptions behind it when giving the answer. He also wondered if AI could help in tax simplification. What would be the implications of writing legislation in a more tech-cognisant way?