Everyone’s talking about AI. Software providers are putting it in every product announcement, and our clients regularly ask how it’s going to change their approach to accounting. Sadly, I’m underwhelmed by the impact of AI on accounting so far. Not because I’m suspicious of new technology. Quite the opposite. I founded Beyond in response to a major technology shift, so I know what tech innovation looks like. For now, AI hasn’t made a dent in this sector. Let’s explore why it falls short.
What a real technology shift looks like
For us, the impact of cloud computing was huge. Cloud-based accounting solutions meant a business owner could access their bookkeeping and accounts from anywhere, and accountants could work on a client’s books without going out to their premises. We were all about cloud from day one and today only take on clients who use platforms like Xero, QuickBooks, Sage, etc.
These new systems could connect to other applications, including the bank, so transactions came in automatically. The client, the bookkeeper, and the accountant could all look at the same system at the same time, which server-based software made very difficult. Things went paperless, and a lot of automation came with it. It really did change everything.
Cloud is the reason Beyond exists. I wouldn’t have founded a practice like ours under the old way of doing things – wading through endless paper records and chasing clients for receipts. I am also very interested in reporting, but getting data out of an accounting system and into clear reports and spreadsheets used to be clunky. With cloud, the data was very accessible, and plenty of third-party apps sprung up to help visualise it.
Another change was to in-house staff. At the start of my career, even a small company of, say, €2 million turnover would have needed a finance team of four: accounting, payables, receivables, and credit control. They’d probably have a senior IT professional and possibly a junior too. You won’t find a dedicated IT person in a company that size now; everything is Software as a Service (SaaS). Accounting has been moving towards outsourcing in a similar way.
Cloud caused a quantum leap in accounting. So how does AI measure up?
What is artificial intelligence anyway?
Let’s be clear about what we’re talking about. People on both sides – the businesses using the software and the companies building it – use ‘AI’ as a catch-all. At least three different things are regularly lumped together:
- Machine learning (ML), which has been around for decades and inside accounting software for many years. Xero’s bank reconciliation suggestions are a good example.
- Large language models (LLMs), which are the tech behind chatbots such as ChatGPT, which arrived in November 2022, or Claude.
- ‘Agentic’ features, the 2026 buzzword, mean that the software takes actions on your behalf rather than just making suggestions.
AI isn’t the first buzzword to promise a transformation; a decade ago the talk was all about the overwhelming power of ‘big data’, but in SME accounting that came to very little. Some of the analytics and automation that already existed are now being lazily labelled as AI.
Lots of promise but little delivery in accounting
At Xero conferences in the mid-to-late 2010s, I heard Xero’s founder, Rod Drury, predict that bank reconciliation would be fully automated within a few years. As I understood it, Xero would draw on transaction data from across the country to accurately predict account codes and VAT rates for each transaction or match it against a bill. It would know, for example, that a payment to Vodafone is always a phone bill at 23% VAT, because every business’s history said so, and code it more accurately than a bookkeeper could.
Xero did move in that direction in 2021, announcing bank reconciliation predictions based on machine learning trained on millions of reconciliations across different organisations. Despite this, bank reconciliation in Xero still takes about as much human work as it did when I sat in those conference halls. It predicts repeat spending within a single business reasonably well, and bank rules are good, but you have to set them up yourself. The automated bank reconciliation I was promised hasn’t arrived.
Where is the intelligent accounting?
Xero’s answer to the AI boom is JAX (Just Ask Xero), described as its agentic platform for small businesses, accountants, and bookkeepers. According to Xero, it reads source documents and extracts the data into the books, and it matches transactions to bank feeds in real time. A good deal of the rest is still on the way: pulling data from bank statements, creating transactions automatically, checking VAT, and handling trickier cases such as splitting a payment between sales and fees.
In practice, nobody I know has found that JAX changes how they work. Much of its capability is announced rather than delivered. Xero has also launched XeroForce, a tool for building custom AI agents with plain-language prompts, but it is currently only open to ‘a small group’ of users with no timeline available for general release in Ireland.
Meanwhile, the cost of Xero has gone up a lot, while the core product hasn’t changed much. The next increase lands on 1st November 2026, and the multi-organisation discount is being removed at the same time. Like so many of the digital services we all use, the AI icing on top isn’t delivering enough to justify the subscription increase.
SMEs don’t have the spending power of enterprises
This is the SME space we’re talking about, and I suspect AI is making a much bigger impact in large enterprises because they have the capacity to put dedicated developers on building AI-based solutions. Whether they call it automation or AI, they have the money to spend and will get the efficiencies in return. Adoption figures point the same way. According to the CSO, 57.7% of large Irish enterprises used AI in 2025, compared with 28.6% of medium and 17.2% of small ones (incidentally, only 4.7% of enterprises used AI for accounting or financial management). SMEs don’t have development teams and must content themselves with the AI features shipped as part of their existing subscriptions.
This year, a company called Andon Labs opened Andon Market, a boutique in San Francisco run by an AI agent. It has real human staff, but the AI chooses the stock, sets the prices, and runs the operation. The backers are covering rent and running costs for three years, so it can learn without having to turn a profit. The early results weren’t great. A few weeks in, Bloomberg reported the shop had lost around US$13,000 and had made basic mistakes like over-ordering stock.
Most Irish SMEs won’t get that luxury. If switching on an AI could make every SME profitable, that would be great. But we’ve seen from the coalface that it’s actually very hard to make an SME profitable. It’s a lot of work to build the service, the trust, the people, the brand, and the culture. An AI can write a plan and give instructions, but that doesn’t mean your employees are going to deliver on it.
There wasn’t much need for what LLMs are good at
Everyday users were bowled over by LLMs, primarily because of how fluently they generate text and how broad their training data was. But fluency isn’t accuracy: LLMs predict the next word, not the right one. That makes them good with language, but less reliable with numbers, and not great with spreadsheets.
Give an LLM your raw data and it will give you a decent narrative about what’s there – like trends, percentages, and comparisons. But that was never a report anyone was looking for before LLMs came along. There was no gap there; it’s just a potentially interesting report that you would never have paid someone to produce.
AI tools now connect directly to Xero. Since May, Claude has been able to answer questions about a business’s live figures: cash position, overdue invoices, and profit trends. In September, Xero added workflows in Claude and ChatGPT that prepare cash flow snapshots, month-end checklists, and lists of overdue invoices to chase. But this is information Xero’s dashboard and reports already provide. Anything an AI produces from a business’s accounts still has to be checked by someone who knows them, so it doesn’t save time either.
Fluent, confident, and possibly wrong
Ask a chatbot such as ChatGPT or Claude what your profit and loss account shows, and you’ll get a fluent, confident answer. Which is also the core issue with LLMs; they are too loose, and answers may or may not be correct. A chatbot can treat a one-off cost as a trend or confuse profit with cash in the bank and sound just as sure when it’s wrong as when it’s right.
The people most likely to ask AI to explain their accounts are the ones least able to tell when the explanation is wrong. If you don’t know the difference between profit and cash, or why a healthy margin can sit alongside an empty bank account, you can’t check what you’re being told. That’s not a small group. In a May 2026 survey of 1,000 Irish adults commissioned by Irish Funds, 44% said they didn’t feel financially literate or weren’t sure.
Relying on AI to read your accounts doesn’t help you build financial knowledge. Every answer arrives ready-made, so there’s no reason to learn what you’re looking at. A good accountant explains the figures in the context of your business, notices when you’ve misunderstood something, and over time teaches you to read them yourself. A chatbot hands you a ready-made answer, and there’s no reason to learn what you’re looking at.
Faster isn’t a transformation
Cloud accounting earned its place by transforming the accounting and bookkeeping functions. Headcounts fell and overheads dropped, while efficiency and visibility skyrocketed. Nobody who took the plunge felt this was anything other than a smart move. AI hasn’t done anything like that for SMEs. The parts that work, such as reading a scanned bill or suggesting an account code, are technology that has been around for years under a new name.
The genuinely new AI features can accomplish simple tasks on instruction, but they save seconds or minutes on jobs that were already quick. If all AI does is shorten tasks that still exist and still need a person to oversee them, I’m not impressed. Cloud accounting didn’t make paper records quicker to handle; it completely abolished them.
AI will only become transformative when its accuracy and accountability reach the point at which the review step can disappear. When processing accounts stops being the biggest part of the job, accountants will be able to focus on directly advising clients. With more of the support they need, SMEs will have a better shot at turning a real and sustainable profit.
Thinking about outsourcing your accountancy function? At Beyond, we don’t just crunch the numbers! We offer impartial business advice and build long-lasting relationships with our clients. Get in touch to find out how we can help!
Rory
P.S. If you find our articles informative, you can add us to your preferred Google search sources.
