I sat down to write something confident about AI and finance, and the first thing I had to admit is that nobody, me included, really knows where it’s going.
Since late 2025 and into early 2026, there’s been a genuine revolution in how businesses use AI. A lot of that, for me, comes down to tools like Claude bringing a proper business focus to the table; not AI as a novelty, but AI as something you can actually put to work. And it’s moving so fast, particularly over the last five or six months, that anyone claiming to be an expert is overselling it.
Here’s the bit worth sitting with: it’s hitting every industry, every sector, every role differently. The nuance of how it affects you; your job, your sector, simply hasn’t washed through from the experts yet. We’re all figuring it out in real time. At Numeric we’ve been active users of Claude since late 2025. For me personally, it started even earlier, a Timoney Leadership Masterclass https://www.timoneyleadership.ie/ back in 2023 got me using AI personally, which turned out to be decent preparation for bringing it into the business.
So with all the caveats firmly in place, here’s how I’m thinking about three things: the systems, the jobs, and the reporting.
The systems aren’t ready — and that’s worth knowing
Most accounting platforms don’t have AI built in. That includes your legacy general ledgers, and yes, it includes cloud systems like Xero too.
This matters more than it sounds. When these tools tell you they’re “working on AI” what’s usually arriving is a workaround, something bolted onto the side rather than built into the core. AI-native accounting platforms are genuinely on the way, and the day they arrive in numbers, the difference will be obvious. Their real advantage won’t be a clever chatbot in the corner; it’ll be the ability to talk to agents and work hand in glove with AI tools.
Legacy systems and let’s be clear, that’s the vast majority of general ledgers in the world, won’t do that. They’re unlikely to change in the short term, for a fairly unglamorous reason: for most businesses, they’re doing a reasonably good job already. Nobody rips out a working ledger for a promise.
On jobs — less drama than the headlines suggest
There’s a lot of fear-mongering about people losing their jobs to AI. Some of it will happen. I’m not going to pretend otherwise.
But the reality is quieter than the headlines. Most roles have parts AI can take over and parts it simply can’t. So the honest picture is that AI makes us more efficient in our roles, it doesn’t necessarily remove the role.
Yes, there’s a version of this where five people are doing the same job, AI removes eighty percent of that workload, and suddenly one person is doing the work of five. That’s real. But it’s rarer than people think. Most jobs are a web of human contact and dozens of small processes and workflows. AI helps with all of it, and removes very little of it wholesale.
The one exception I’d flag: if a significant part of your job lives in Excel, it’s quite likely your role changes dramatically, and soon. Our own experience at Numeric is that a lot of business workflows will be slow to evolve, business is clunky by nature, but anywhere the work is analysis on data, AI will be doing that very soon.
On reporting — where the caution lives
Reporting is close to my heart, so forgive me a slightly longer thought here.
Numeric has always been focused on standardised reports, built with SaaS tools, for three reasons: so reporting is consistent, efficient, and, most importantly, easy to validate. The whole point is that time isn’t spent checking and rechecking the same information.
Now, AI can produce exceptionally good reports genuinely good, with real context, real colour, pulling data from many different sources at once. That’s the upside, and it’s a big one.
The problem that remains is validation. AI behaves a bit like a person gathering information from everywhere at once, and it can trip itself up matching that information and keeping it relevant. Which means the discipline doesn’t go away, it changes shape. We use AI tools to validate information, and we use humans to read and check it. That second part is expensive in time, and it needs treating with caution rather than waved through because the output looks polished.
So we still favour a proper monthly close: P&L, balance sheet, cash flow and forecast, coupled with your KPIs and the operational and commercial story around them. And all of it drawn from robust sources, the general ledger, operational and commercial dashboards.
The question I’d leave you with
None of this is a reason to sit on your hands. It’s a reason to be specific.
So here’s what I’d ask you to think about: do you actually know which parts of your role AI is about to change – and which parts it never will? Because the people who answer that well over the next year won’t be the ones who feared it, or the ones who believed every headline. They’ll be the ones who got specific about their own numbers.