AI 'slop' in the books: The rising cost of fixing chatbot errors (from a British trade magazine for accountants)

[highplainsdem]

This is from Accountancy Age - https://en.wikipedia.org/wiki/Accountancy_Age - and what it says about AI for accounting is pretty damning.

https://accountancyage.com/2026/01/08/ai-slop-in-the-books-the-rising-cost-of-fixing-chatbot-errors/

AI 'slop' in the books: The rising cost of fixing chatbot errors
As "AI slop" invades the ledger, a new Dext study reveals that half of UK firms have already suffered financial losses due to rogue chatbot advice. With 2026 looming as a potential breaking point for business failures, we examine why the profession is making an urgent call for regulation.

Nikita Alexander
January 8, 2026


-snip-

Top AI-Generated Errors Reported by Accountants:
Incorrect Business Expenses: 46%
VAT Miscalculations: 41%
Flawed Personal Tax Planning: 35%
Payroll & Business Tax Errors: 34%

-snip-

While AI is promised as a time-saver, 93% of accountants now spend significant portions of their month fixing AI-driven mistakes. Nearly 40% of practitioners are losing between 4 and 10 hours per week unravelling incorrect data.

-snip-

Perhaps most concerning for the long-term health of the profession is the shift in client attitudes. The research notes that 72% of accountants have seen an increase in clients using AI outputs to actively challenge professional advice.

We are entering a cycle where “The AI said X” is being used to justify aggressive or fraudulent tax positions. This puts practitioners in a difficult spot, balancing the role of a trusted advisor with the need to act as a “de-programmer” for clients who believe their £20-a-month subscription knows more than a Chartered Accountant.

-snip-

Great.

And keep in mind that thanks to the AI bros and our dunce of a president, the US government is now running on chatbots, and individuals are pressured by tech companies to use them almost everywhere.

I'll admit I laughed at the line about the people using chatbots needing to be deprogrammed. But it is a serious problem - people are too susceptible to chatbots, even when what the chatbots are saying is ridiculous and even harmful.

[UpInArms]

1. I have a friend who

Refuses to use ai with their job function, as it makes too many errors and requires too much time to repair them

[highplainsdem]

2. Good for your friend! I've always suspected that most people who believe AI saves them a lot of time

simply aren't catching and fixing the errors.