AI in practice

Keogh highlights ‘dynamic scenario modelling in financial planning’ as one of the big wins. ‘It’s a task that in the past would have taken me days. Today, I can feed in key variables and AI generates multiple, detailed forecasts in just minutes.’

At the more routine end of the spectrum, she identifies another game-changer: bank reconciliation and variance analysis. ‘AI has dramatically improved efficiency, reduced human error and freed up time for strategic activities. This shift not only saves resources but also enhances audit readiness and compliance.

Keogh recognises that getting the best from AI requires a considered approach – ‘breaking tasks into smaller pieces. Give it too much at once, and it risks glossing over key details, which in finance can’t happen.’ However, even this limitation has upsides, as she explains: ‘The discipline of structuring inputs improves the quality of my own thinking.’

The key is not to confuse convenience with complacency. ‘AI is generally strong with calculations and trend analysis but can sometimes misinterpret context such as industry-specific nuances. I use it confidently for computations, reconciliations and statistical analysis but always validate final outputs,’ she says.

Keogh recommends listening to The Diary of A CEO podcast’s interview with Mo Gawdat, the ex-Google executive. ‘He argues that within three years the way we work will completely transform,’ she says.

The online accountant’s view: ‘have the right tools’

For online accountant Clive Hand FCCA, owner/manager of Bluestone Accounting, the year has been about ‘using more AI tools, like Microsoft Copilot, Loom AI and Descript. It’s not about having one AI tool; it’s about having the right collection of AI tools,’ he says.

He advises that any use of AI for number crunching should involve a manual review, with the suggestion to include the instruction to ‘show your calculation’ in any AI prompt that involves figures.

‘Prompt engineering is really important if you want to maximise the quality of the output. For example, a prompt might look like: “Act as an Irish chartered tax adviser. Question: [enter text here]. Client facts: [enter bullet list]. Jurisdiction: Ireland; quote Revenue manuals/Tax Acts with section numbers and short quotes. Deliverables: (1) assumptions, (2) analysis with alternatives, (3) risks/edge cases, (4) actions/checklist for ROS, (5) prepare a client-ready email (200 words), (6) unknowns – ask specific follow-ups.

‘For any calculation, show workings and a cross-check. If information is missing, don’t guess, state the decision tree.” Remember never to use your clients’ actual details in ChatGPT useless you are using ChatGPT Team,’ Hand explains.

Hand reports conversations with clients around AI have also increased over the year: ‘It’s primarily around how they can use it for their own business, mainly focused on systemising routine bookkeeping tasks and how to automate these tasks so they can focus on more value-add activities. We also discuss how AI can be used for activities like credit control and chasing customer payments to help create a positive cashflow impact.’

Hand expects that the introduction of AI agents, or systems that can act autonomously on a user’s behalf to complete goals, will see usage grow in 2026. ‘The fact that we can speak with AI agents in a natural language will make them extremely accessible to everyone’, he says.

The entrepreneur’s view: ‘only as good as the data’

For entrepreneur Noel Lourdes FCCA, AI is front and centre in his stated ambition to transform how we work in business. ‘An area of focus is “cognitive archiving” to solve the problem of  recruiting and retaining people in certain jobs,’ he explains. ‘For example, in the sourcing of aircraft parts, traditionally the person who does this job has 20 to 30 years’ experience, learnt on the job. When they retire, all the knowledge goes with them.

‘We think AI can be used to temporarily fill in certain jobs while you recruit their replacement, using an experienced person’s knowhow to show how to get things done.’

There are also opportunities for transforming the everyday work of accountants.  ‘What was previously laborious does not mean previously right. For example, in calculating product margin, it may have been laborious to manually take the job costing data on the materials and labour hours to manufacture the product. But this laborious effort failed to factor in other issues. AI could help unearth intelligence that is not obvious to an accountant,’ Lourdes says.

Lourdes identifies a difference in approach for SMEs and larger organisations.  ‘Large corporations can afford the spend required in getting good data scientists and AI engineers to make it work for them. SMEs don’t have the IT people and technical resources to properly use AI and hope that AI will just work outside of the box – which is what we see as the main driver of AI usage in 2025, led by ChatGPT.’

On the question of future work displacement in the profession, Lourdes is clearheaded. ‘The next generation are not interested in roles that are laborious and not “glamorous” such as back-office tasks like procurement.  If the next generation don’t want to take on these roles, leave them to AI.’

The managing partner’s view: ‘in some areas, invaluable’

Michael Queally FCCA, managing partner of AMQ Accountants & Auditors, says AI has become increasingly useful . ‘We now use AI to assist with drafting work papers, client letters and drafting code for VBA/Python to automate repetitive tasks,’ he says.

Some limitations that were in evidence early remain concerns: ‘I found earlier versions of ChatGPT tended to get confused in certain projects if they went on too long,’ he reflects, adding: ‘I don’t rely on AI to analyse numbers at the moment.’

An open, flexible approach to AI within the firm is matched by a sense of vigilance on the outputs. ‘Employees are free to use AI where they believe it will add value. Every AI output is reviewed by a human and, where relevant, checked against primary sources such as Revenue, CRO or the Irish Statute Book,’ he says.

Queally has witnessed some of AI’s infamous hallucinations, but the grounded approach means they haven’t been caught out by them. ‘I once asked ChatGPT a tax question and it gave me a perfect answer – if I was looking for UK tax advice,’ he recalls.

He also believes AI will support, not displace, the next generation of accountants. ‘We will continue to have a significant need for graduates,’ he says. ‘Their future roles will be more challenging and fulfilling than they had been in the past as the tasks they will be doing will be at a higher level.’