Live data from Hacker News

Tell HN: AI coding is sexy, but accounting is the real low-hanging target

news.ycombinator.com

51–58 of 58 posts

Re: Tell HN: AI coding is sexy, but accounting is the real low-hanging target

#51
post #49
post #47

Earlier quoted context omitted.

They might not be solvable but you can get 5-10% Improvement on them, unfortunately you can't do a new product that is exactly like QuickBooks but 5% better at reconciliation etc.

LLMs by their inherent nature cannot be relied on to be true and correct, which by coincidence are the only traits that matter in accounting. If you want better software, then sure, maybe a coding assistant can help you write it faster, but when it comes to actually doing accounting I would not rely on an LLM in any way shape or form any more than I would do so for law.

Bingo! You found the prize! Putting tech that is prone to hallucination in charge of anything that has serious consequences when it's wrong is a terrible idea. You do not want hallucinated payments or receipts, or legal citations. You want these things to be both true and correct, EVERY TIME.

Re: Tell HN: AI coding is sexy, but accounting is the real low-hanging target

#52
post #12

Earlier quoted context omitted.

I disagree on this, there are plenty of problems in accounting that an LLM can help with. I’ve built some software[0] that analyses general ledgers and uses LLMs to call out any compliance issues by looking at transaction and account descriptions. Is it perfect, nope. But it’s a hell of a lot better than sifting through thousands of transactions manually which accountants do and get wrong all the time. [0] - https://…

> I disagree on this How can you disagree with the fact? Some specific examples (like the one you mentioned, _adjacent_ to accounting per se) don't disprove the main point that 100% accuracy is fundamentally impossible with LLMs, while critical for all key accounting aspects.

I think we misunderstand each other.

Accounting is quite broad and there are many examples of where LLMs can help.

For example, what is tax deductible is subjective to local laws. This is essentially a classification problem that LLMs are particularly good at.

Re: Tell HN: AI coding is sexy, but accounting is the real low-hanging target

#53

"It’s verifiable. The books either balance or they don’t. Ledgers either reconcile or they don’t. There’s almost always a “ground truth” to compare against (bank feeds, statements, prior periods). It’s boring and repetitive. Same vendors, same categories, same patterns every month. Humans hate this work. Software loves it." These are all true statements, but all of those things are solvable with classic software. Qui…

I go the other way and say this is a vast oversimplication of the job by a tech bro doing what tech bros typically do.

I am not an accountant but for many years I worked adjacent to them as a developer. I got a lot of time to ask questions and I was generally curious. Even at the SMB level accountants don't necessarily always have a "ground truth". There are so many ways to bury financial data that it needs almost constant vigilance. Yes, in theory, there is one ground truth (inputs should balance with whats there) but in practice humans are SHOCKINGLY good at committing accounting fraud. GAAP is another thing.

Re: Tell HN: AI coding is sexy, but accounting is the real low-hanging target

#56
post #11

Lots of this is already being done (and using computers to check books balanced predates latest gen AI by some decades). But of course what accountants actually get paid for is tax strategy, edge cases, messy history and talking to authorities (or at least having their stamp of approval on it if the authorities come calling). There's plenty of market for writing software to automate aspects of invoice reconciliation…

At the SMB scale accountants are mostly paid to coach/pester/goade the employees to hand in the necessary paperwork in time. The accountants job is relatively quick from there on.

Taking a corporate accounting course was required for my degree. It was also a fantastic investment because I’ve managed to avoid needing to hire an accountant or bookkeeper to setup or maintain my books. I’ve migrated between multiple accounting systems on my own just fine, starting from posting migratory opening entries manually.

From there I just had to create a few rules and add a custom reconciliation model for things like the cash back card my company uses. There’s some data entry but it’s all extremely straightforward. Not really a job for AI in the best of times.

What was particularly useful was being able to map my revenue and expense cycle onto accounts that actually make sense for my business and maintain compliant reporting.

I just have a CPA who specializes in taxes and is an EA with the IRS do my return for me each year for only about twice the cost of doing it myself with another application. I find that totally worth it because she can use decades of experience to inform advice that she provides.

And I don’t even have to wonder if she is confidently incorrect.

Re: Tell HN: AI coding is sexy, but accounting is the real low-hanging target

#57
post #49

Earlier quoted context omitted.

LLMs by their inherent nature cannot be relied on to be true and correct, which by coincidence are the only traits that matter in accounting. If you want better software, then sure, maybe a coding assistant can help you write it faster, but when it comes to actually doing accounting I would not rely on an LLM in any way shape or form any more than I would do so for law.

Bingo! You found the prize! Putting tech that is prone to hallucination in charge of anything that has serious consequences when it's wrong is a terrible idea. You do not want hallucinated payments or receipts, or legal citations. You want these things to be both true and correct, EVERY TIME.

“We had to re-state our financials and amend our taxes because the AI screwed up and we didn’t have anyone who understood accounting look at our books.”

Re: Tell HN: AI coding is sexy, but accounting is the real low-hanging target

#58
I find it interesting that tech people think LLM's will automate most jobs other than software and other creative jobs. Its hard to imagine for us techies that billions if not trillions of investment is mostly targeting disrupting software engineering, artists, etc as the low hanging fruit. The industry mostly disrupted itself and gave a lesson to other industries as to what not to do.

In the end AI will take only the "fun jobs" initially -> the ones where the barrier was creativity and intelligence. Ironically all the boring, non fun and challenging/accountable tasks will remain. All those "auxiliary" jobs in tech firms and those corporate workers who mostly did admin and added little value while you pushed novel products will most likely remain somehow -> they were never judged/evaluated on value delivered so there's nothing to disrupt.

Post reply on HN