Every few years, a new technology arrives promising to replace the accountant. Spreadsheets didn't. Cloud software didn't. AI won't either — but it is genuinely changing what accountants spend their time on, and the firms getting the most value from it share one thing in common: they didn't try to remove the human, they repositioned them.
The false choice
Framing this as "AI vs. traditional accounting" misses what's actually happening. It's not a competition between a model and a person — it's a redistribution of work. AI takes on the high-volume, repetitive, pattern-matching tasks. The accountant takes on everything that requires context: why a client's margins moved, whether a transaction reflects real economic substance, how a regulator is likely to view a judgment call.
What thirty years of practice adds that a model can't
Pattern recognition built from data is powerful, but it's bounded by the data it's seen. An accountant who has sat across the table from hundreds of business owners going through a cash crunch, a due diligence process, or a tax audit brings something a model trained on transaction data alone doesn't have: judgment shaped by consequence. That's the layer we've never automated, and don't intend to.
Where the balance actually lands
In our own practice, roughly 60% of routine transaction processing is now AI-assisted. None of our audit opinions, tax positions, or advisory recommendations are. That split isn't arbitrary — it reflects where automation reduces error and where it introduces risk that outweighs the time saved.
The SMEs that get this balance right treat AI as a force multiplier for their finance function, not a replacement for it. They still have a named accountant who understands their business. That accountant simply spends less time on data entry and more time on the decisions that actually move the business forward.
Our approach
We built our AI-powered automation the same way we built our audit practice thirty years ago: conservatively, with a person accountable for every output, and only after testing it against real client data. If you're weighing how much of your own finance function could be automated without losing the judgment that keeps you compliant, that's the exact conversation we have with SMEs every week.
