AI won’t replace accountants, but inertia might

Every AI conversation tends to follow the same path. First the demos, then the hot takes, then the LinkedIn posts about “the future of accounting.” And then, mostly, nothing changes. The firms don’t move.

These groups have focused on what AI actually looks like inside a working practice, not on a vendor’s slide. Two things stood out: nobody needed convincing that AI matters, and almost nobody could tell me what to actually do about it or where to start.

That gap isn’t about cost, and it isn’t about scepticism. It’s that most accountants don’t have a trusted source of truth telling them which tools are genuinely fit for their work, which are hype, and which will still exist in two years. In a profession built on getting the detail right, that uncertainty is enough to stall a decision. Firms would rather do nothing than do the wrong thing.

Which makes it more interesting that the firms actually making progress are, almost without exception, not the ones running a structured, top-down AI program.

They’re the ones where someone – often a younger team member with a genuine interest in the tools – was given room to explore, test something on a real piece of work, and show the practice what’s possible. The adoption spreads from that demonstrated use case, not from a steering committee memo. There’s a lesson in that. Firms don't need a transformation strategy to get started. They need to notice who on their team is already curious and get out of their way.

So, what actually changes first? The high-volume, low-judgement work: document classification, data extraction, first-pass checks against known rules. This is the work that currently consumes the bulk of a junior accountant’s week without drawing on much of their actual training. It’s also the work clients never valued in the first place. Nobody pays a professional fee because they enjoyed the invoice reconciliation.

What doesn’t disappear – and becomes more valuable – is judgement. AI can surface an anomaly, a risk, a pattern across a client’s numbers. It cannot decide what that pattern means for this client, in this industry, at this point in their life or business. It cannot sit across the table and have the conversation that follows. As the mechanical work is absorbed elsewhere, the accountant’s time gets pushed toward exactly what clients actually pay for: interpretation, advice and trust.

Here’s what I’d actually be doing about it:

Stop looking for the perfect platform. There isn’t one, and waiting for a trusted, comprehensive verdict on “the best AI tool for accountants” is itself a form of inaction. Pick one narrow, painful, repetitive task and properly test a tool against it.

Find your curious person. Every practice has one – someone already using AI tools in their own time, quietly, because they find them interesting. Give them a real file and permission to try. The firms winning right now didn’t get there through a committee.

Don’t hand judgement to the machine. The audit, the risk rating, the advice conversation – that stays human and should be seen as such. The efficiency case for AI is strongest exactly where the judgement case is weakest.

Talk to your team about what their week actually looks like. If nobody can name the three most repetitive, lowest-value tasks eating their time, that's the first problem to solve, before any tool gets chosen.

By 2030, I expect the practice model to look genuinely different: smaller technical teams doing compliance work, supported by AI rather than headcount; larger advisory functions, because the capacity freed up by automation has to go somewhere, and client demand for advice on structure, tax and strategy isn’t going away; and a widening gap between firms that treated this as an efficiency project and firms that treated it as a chance to change what they actually sell.

I don’t think this is a story about job losses. I think it’s a story about which firms decide to become something more than a compliance function, while they still have the choice.

The accountants I speak to aren’t worried about AI replacing them. They’re worried about being left behind by firms that move first.

They’re right to be. AI isn’t the danger. Standing still is.

John Munden is chief strategy officer at Cloudoffis.