Close Menu
    Technik

    Goldman Sachs beauftragt Claude von Anthropic mit der Automatisierung von Buchhaltungs- und Compliance-Rollen

    Share.
    View 27 Comments

    27 Kommentare

    2. Secure-Address4385 on

      Not that surprising accounting and compliance are rule-heavy and document-driven. The bigger question is whether AI replaces human oversight or just pushes people into review roles.

    3. CurveSudden1104 on

      So let’s actually discuss the technology not fear monger for a minute.

      The fact that Anthropic is gaining this much traction in business has to be freaking OpenAI the fuck out. I wonder if this is why Altman has been crashing out on Twitter lately.

    4. Friendly_Budget_3947 on

      Using large LANGUAGE models to solve MATH problems probably has interesting results 😝

    5. 7___7 on

      AI signed off for this material error, so the executives aren’t culpable but the laptop over there is. Go ahead and send the laptop to 25 years of jail, we’ll have a pizza party for staff to grieve, and the CEO can focus on which yacht they can add to their portfolio from their upcoming quarterly bonus.

    7. mtranda on

      I was wondering about half a year ago how long it’ll take before someone comes up with this unbelievably idiotic idea. The answer was six months apparently.

      Accounting is fully deterministic. It’s a very, very strict matter of „take column A and multiply it by column B“. Every single time. And guess what? So is the accounting software that has been successfully doing this job for the past nearly 100 years.

      Do you know what is terrible at being deterministic? A statistical engine incapable of providing two identical answers in a row given the same identical inputs.

      Nobody wants an accounting system that has a chance to „get creative“.

      If you want a demonstration of just how badly this can go, just look up videos of AI playing chess. 

      This is classic example of a solution looking for problem. 

    8. user284388273 on

      I asked Claude to analyse our company’s code base to identify the cause of a bug, it spent a lot of time thinking and replied with a brilliantly confident answer pinpointing the exact cause…

      Only one problem, it was completely wrong…

    9. Expensive_Shallot_78 on

      I am using Claude, Claude Code Opus, and Gemini Pro for over a year now and it does monumental bullshit every other minute. How?

    14. Moist1981 on

      I suspect they’re using AI to improve the efficiency of the process rather than to outright replace workers. AI is being used at various firms to bring up associated records etc. I don’t think they’re yet going to be using it for expert judgement in accounting matters (it will be fun when that dies eventually happen and inevitably at some point will suggest a course of action that regulators disagree with).

    16. Middleage_dad on

      Step 1:  Setup AI to handle accounting
      Step 2:  Do a bunch of shady accounting shit
      Step 3:  Blame the AI if you get caught. 

    17. Disgruntled-Cacti on

      “Goldman could next develop agents for tasks like employee surveillance or making investment banking pitchbooks, he said. “

      I’m glad we got the dystopian future of mass surveillance and unemployment rather than the utopian future abundance and having menial tasks like laundry done for us. Feels really great.

    18. Few_Cauliflower2069 on

      Sounds like somebody really wants a visit from the IRS and some regulative authorities. That’s yet another bad idea for the use of ai

    19. Jhonka86 on

      This is an insane fucking choice. GS is walking dick-first into a blender made of lawsuits and is bragging about it.

      AI lies. Like, all the time. People believe that it’s trained on accuracy, but in reality it’s trained on believability. Each time an association passes without objection, it’s reinforced.

      It’s essentially a giant gaslighting / Mandela Effect generator. The more times a narrative is pushed and believed, the more it pushes the narrative. But as soon as you call it out on being wrong, it bends over backwards to praise you for being so smart and wise and right to challenge it.

      There’s a group of humans that act like this. We call them psychopaths.

    20. whinner on

      I can’t imagine a CFO signing off on an AI created financial statement when their own ass is on the line

    21. stuffitystuff on

      I mean I assume/hope that they’ll be using Claude to write software to automated accounting and compliance roles not actually replace people.

      I’ve written the same software before the advent of LLMs and freeing people from that dreck is sweet, sweet release. It’s probably stuff like „take a screenshot of this machine failing to ping a machine not on the corporate network“ and that sort of thing.

    24. Modroidz on

      How will they program in loopholes when they wanna do shady deals and the AI says it goes against protocol?

    26. AndyTheSane on

      I preferred it when they were handing out mortgages for any amount to anyone who could fog a mirror.

    27. abc13680 on

      It’s interesting to see the comments here in a “technology” sub. I’m assuming a large portion here only interact with chatbots or off the shelf code assistants.

      GS probably has thousands of devs, thousands working on accounting and compliance, reams of strict training doc and compliance records, and millions of spent annually of accounting software.

      This is not asking Claude to do accounting. It’s building agentic infrastructure to allow a language model to transform written text (unstructured data) then use pre specified tooling to hook into existing software. The LLM isn’t doing accounting math (in much the same way most accountants are doing a ton of accounting math , they offload to software).

    Leave A Reply