Built AI forensic accounting software with my dad
case-trail.com
case-trail.com
There is something very concerning about this article: submitting private information to LLMs w/no privacy guarantees is probably a crime. I strongly recommend taking down this article and that you stop submitting private information to LLMs with no privacy guarantees until you have spoken with an informed lawyer on this matter.
Local models may be of assistance here, but you need to be very careful.
Also, your comment is just trivially unininteresting to begin with: what is "submitting," what is "private," what is a "privacy guarantee," what is a "crime"? It's just outrage slop. If you're a lawyer and have something substantive to say, please do so (and a lot of lawyers on HN do), but otherwise, Lincoln put it best: "Tis better to be silent and be thought a fool, than to speak and remove all doubt."
Who are you (or who am I) to decide that? The entire point of a show HN is to be non-judgmental and charitable, otherwise it's just going to turn into a cynical echo-chamber. The famous Dropbox comment is a cautionary tale for a reason.
> Person A: yo wtf is wrong with you
> Person B: Who are you (or who am I) to decide that? The entire point of a show HN is to be non-judgmental and charitable, otherwise it's just going to turn into a cynical echo-chamber. The famous Dropbox comment is a cautionary tale for a reason.
I'm not trying to gatekeep and say that only domain experts should be allowed to build software, but part of being an engineer is doing due diligence to understand the domain well enough to build the product. If OP failed to recognize that any forensic accountant that gets caught uploading privileged documents into a random AI tool would be both breaking the chain of possession of that document AND client privilege, what other mistakes did they make along the way?
I went through the entire website and couldn't find a single mention about privacy. I'm not a domain expert, but I would expect the product site for any legal tool to at least have a disclaimer.
Regardless, I'm comfortable being a called fool.
Unless OP is using hosted models, especially those with always-on training, that's quite clear cut breaking at least privacy laws, likely more, especially if the court documents are additionally protected.
So that's basically showing the HN how egregiously a number of lawyers, accountants and paralegals "conspire" to break the law in order to process more cases in parallel and earn more money.
I think that's pretty accurate?
If OPs father doesn't want to do it manually they must at least run it locally, or obtain the court permission to share the privileged information with a number of third parties, possibly shoving it into the future corpus of information.
I'm curious about the 60% automation of financial/forensic analysis - what's missing? Is it stuff that's purely blocked by model capabilities, or are there places where scaffolding is likely to bridge the gaps?
Also curious about the workflow - is this more individual, LLM-driven features or agentic workflows? Looked like the former from the product video but there wasn't a ton of UX shown there.
I ask largely because this seems like the sort of thing where you could really start to string these features together in such a way that you start with a description of the case and whatever files you have, and then an agent does its analysis of the docs, spins up action items (get missing docs, confirm that X ambiguous doc is what the AI characterized it as, etc.) and tracks the progress of all of them, leaving your forensic accountant there in a supervisory role, managing and providing expertise.
It feels like that's the way a lot of expert analysis jobs like this are headed. I've been working on the same sort of flow to use agents to manage my business. Started with LLM skills that could be used to handle tasks I used to do myself, and since then I've increasingly been having AI use those skills on its own without me invoking them and chain things together into full blown workflows. Some parts I'm still supervising closely, but others that have been working consistently for a while I now don't really watch unless Claude flags something for me to review on my dashboard.
First of all accounting as a whole is incredibly broad. The fact you don’t recognise that in your post with nuance shows you have zero clue what you are talking about.
E.g llm’s are useless in tax auditing. How do I know this? My brother is a partner at pwc.
Wanna talk about deferred tax? Let’s go.
I’ll wait… I bet you know fck all about finance and accounting and your knowledge is a mish mash of stuff you’ve read from a google search.
Junior-level ICs are on the chopping block now, and senior ICs + middle management are next
Here is my hot take. AI is going to replace some developers (not all) and the first ones it replaces will be the ones who can't code without it. The developer in this story provided a relationship with a forensic accountant, a few discussions with paralegals, and limited guidance to an agent. The agent did literally everything else, including writing the article!
Because most AI hypers have extremely low standards for any form of text - be it code or prose. If one is to believe code doesn't matter, then why would would prose matter either?