I own a couple of small businesses, and I've tried a few things with my books - was on Bench for a year (thankfully not the year they shut down, and they were so incompetent I dropped them before that), tried to do them myself for a year, then upon realizing the P&L generated did not match the numbers I expected, hired bookkeepers on Upwork to fix them for me.
I really feel like I ought to be able to do them myself - it's just following some rules, and my accounts aren't that complex. Still, it was just enough of a pain that it was easier to hire someone overseas for cheap (especially since I know what the business' numbers should roughly come out to, so I can validate their work).
But as I've been using the latest AI models, I really feel like this is something that's going to be fully automated by AI in the next 1-2 years (at least for my very simple use case). It's simple enough that an AI agent should pretty trivially be able to fetch the docs from the various places that I sell upload them, categorize transactions (this is already basically automated by rules for me anyway) and then do whatever it is that bookkeepers do to close the books.
I can't help but think that bookkeeper is not going to be a profession in five years, and I'm just not sure where those people go. It's not like automating bookkeeping will expand the economic pie enough to create new jobs - I don't believe it's a bottleneck to anything at this point.
Many customers have asked me about AI offerings, and I am considering them. While this is doable with modern LLM technologies, I need to consider many issues.
The first is that nobody, myself included, likes their data being part of someone else's machine-learning training pipeline. That's why I promised my users that I wouldn't use their data for machine learning training without asking for explicit consent (and, of course, anonymization will be needed).
While I know everything involved in AI sounds cool, do we really need LLM for a task like this? Maybe a rule-based import engine could kill 95% of the repeating transactions? And that's why I built beanhub-import[1] in the first place. Then, here comes another question: Should I make LLM generate the rule for you or generate the final transactions directly?
Yet another question is, everybody/every company's book is different from one to another. Even if you can train a big model to deal with the most common approaches, the outcome may not be what you really need. So, I am thinking about possibly using your own Git history as a source of training data to teach machine learning models to generate transactions like you would do. That would be yet another interesting blog post, I guess if I actually built a prototype or really made it a feature for BeanHub. But for now, it's still an idea.
I really feel like I ought to be able to do them myself - it's just following some rules, and my accounts aren't that complex. Still, it was just enough of a pain that it was easier to hire someone overseas for cheap (especially since I know what the business' numbers should roughly come out to, so I can validate their work).
But as I've been using the latest AI models, I really feel like this is something that's going to be fully automated by AI in the next 1-2 years (at least for my very simple use case). It's simple enough that an AI agent should pretty trivially be able to fetch the docs from the various places that I sell upload them, categorize transactions (this is already basically automated by rules for me anyway) and then do whatever it is that bookkeepers do to close the books.
I can't help but think that bookkeeper is not going to be a profession in five years, and I'm just not sure where those people go. It's not like automating bookkeeping will expand the economic pie enough to create new jobs - I don't believe it's a bottleneck to anything at this point.