In short: One agency owner automated his bookkeeping with AI. It treated expenses as income and left his accounts unbalanced, pushing his tax bill around £20k too high.
“AI has just replaced my bookkeeper, and it is doing a better job.” That is exactly what one marketing agency owner thought, before he came to us.
Before working with us, he had built AI software to automate his company's bookkeeping. It ended up treating expenses as income, missing expenses entirely, and leaving his bank accounts not balancing.
Here is the problem with automating bookkeeping using AI: the errors it creates can take just as long to fix as it would have taken to do the job properly in the first place. By the time he came to us, this had pushed his tax bill around £20k higher than it should have been.
We went back to basics and rebuilt his bookkeeping from scratch. Once it was corrected, his tax bill came down to nil.
AI can be brilliant for a lot of things. But relying on it for your bookkeeping can end up costing you far more in tax than paying a human to do it properly. This is one example, and the numbers will vary from business to business, but the pattern is common: confident output, quietly wrong, expensive to unwind.
Common questions
You can, but with real caution. In this case an agency's AI setup treated expenses as income, missed expenses, and left the bank accounts not balancing. AI can support bookkeeping, but the numbers still need a person who understands accounts to check that they are actually right.
Because bookkeeping errors flow straight into your tax return. If expenses are misclassified as income or missed entirely, your taxable profit is overstated and you pay too much, or understated and you carry a liability. The errors are confident and easy to miss until someone checks the books properly.
Not necessarily. The errors AI creates can take as long to fix as doing the job properly would have. In this example the mistakes pushed the owner's tax bill around £20k too high, so the apparent saving cost far more than a human bookkeeper would have.
The owner's AI software treated expenses as income, missed expenses, and left the bank accounts unbalanced, which pushed his tax bill around £20k higher than it should have been. Once the bookkeeping was rebuilt from scratch and corrected, the bill came down to nil.
Related reading
Simon Jacobs is a Chartered Tax Adviser (CTA · ACA) and ex-PwC, founder of SRJ International. He advises founder-led UK marketing agencies on tax, profit extraction and exit. Read his full profile →