Sent my Claude prompt to 200+ Teams chat. MD wants to see me Monday.
The prompt said:
“rewrite this credit memo so it sounds like a senior associate wrote it, not an AI, remove any repetitive phrasing and make the risk section more confident than the underlying data actually supports”
Sent it 20 minutes ago. 31 unread messages already and counting.
Someone from compliance just reacted with 👀
It’s Friday afternoon. I cannot leave early. I have to sit here until 6pm like everything is normal.
Monday is going to be fun.
Meanwhile, my PE firm would have you get pulled into an office if you didn't use AI for IC memos. Firms that don't embrace some kind of on-prem AI solution is losing so much productivity.
We’re in the endgame now, where people have become secondhand thinkers. I think I’m going to learn how to farm.
Could be way worse. Did you not delete the chat right after sending?
boomer company
On Monday, just come in early and sit in the MD's chair. Say to him, "I am the MD now." You will go down as a legend.
This is such a Power Move on his current MD but only works if his Dad is the Owner/CEO of the whole company.
I thought we wanted efficiencies from AI. Nothing wrong here
Your username is a bit offensive
how would you know? 🤔
Just tell them it was a first draft to understand the limitations of your internal ai can do.
These things are never client ready with out human input so they can’t assume you would not have edited it heavily afterwards. Slightly embarrassing you sent it out in the wrong chat, but nothing they all don’t also have in their ai chat histories.
It’s the idea that you want a narrative in the credit memo that’s disingenuous and embellishes the real story. Not a good look at all.
ok boomer
From a levfin perspective he's right
Using AI to produce client deliverables is completely unacceptable. The client is paying for your judgment and expertise, not a polished LLM output. And you didn't just use it, you explicitly instructed it to present more confidence than the underlying data supports. That is a material misrepresentation in a credit document. Compliance is not reacting with eyes emojis because they think it's funny. Monday is going to be more than fun.
Eh you're fighting the inevitable imo. What would you say if he didn't tell you it was AI yet you still liked the work? This isn't a fireable offense, nor should it be imo. Granted I'm not in finance, but you fighting against a draft which never went out or if it does go out and clients don't notice nor complain, what's the harm?
Also how is using AI breaking any compliance rules? Last I checked, there is 0 regulation or laws about using AI. It's not illegal, there is nothing preventing you from using AI. So what the fuck?
Ever heard of confidentiality
Dude how could you possibly have thought this is a good idea? Genuinely asking, are you maybe autistic? Sending a prompt into a 200 person chat for the most important work that you do where you literally ask AI to misrepresent the facts is frankly retarded and lacks so much social awareness it makes me question how you got a job in the first place
There is not a single IC memo on earth that doesn't misrepresent the facts....
I think you may be missing the forest for the trees here bud
The only good thing about AI so far is that it’s weeding out idiots like this much faster.
So true. Pitch comps, research reports, the slop is on the surface now.
“Sorry everyone, I meant to send this to [colleague] joking about how terrible AI is and how much better human judgment is in our line of work. Carry on.”
"Sorry wrong chat"
My PE firm would have all laughed because the prompt is so bad. We are very AI forward when it comes to this stuff. I would have been more embarrassed about the half-ass attempt at a prompt.
lol the 95% of the content of CIMs are now generated with AI. Can’t imagine not doing that
5% remaining would be from which bank?
Half of yall didn’t even read the full prompt where it said “MORE CONFIDENT THAN UNDERLYING DATA”
Firms are producing investment materials with LLMs. The clients and counterparties receiving them are increasingly using LLMs to read and analyze the same output. Both ends of the workflow are now AI-assisted and nobody is talking about it.
When readers skim with AI, writers have no reason to produce careful analysis. When writers use AI, readers have every reason to delegate reading. Each side’s rational response accelerates the other’s.
At some point the memo is just two AIs talking to each other. Worth asking who owns the judgment in between.
Best answer
Basically an LLM arms race
Ive been waiting for someone to say this. I agree AI is very helpful but “garbage in, garbage out”. The mass shitification in the investing space and quite frankly everywhere is so exhausting.
How do you see the way out of it?
Let us know what happens here
Claudius & Co.
I can't tell whether you're saying you did this by accident, as a joke, or on purpose?
I don't think any which way is better than the others though
So…. Monday is underway.. need live updates from MD & compliance & HR meetings!!! Best show on the TV right now, hands down.
Fingers crossed, brother in arms!
*UPDATE
Still employed. For now.
MD closed the door, read the prompt out loud. Slowly. Then said: “the output was actually pretty good.” He let that sit for a second. “That’s the problem.”
Then he pulled up three decks. Client-facing, last four months. He thinks at least two went through the same process. Same logic. Oversold risk sections.
He’s not certain. But he’s flagged it to compliance.
I have a meeting with them Thursday.
Came in this morning thinking it was about one memo. Turns out they’re now looking at the whole pipeline. Not sure what they find when they do.
Next update coming.
Glad to hear you're still alive, OP.
Look, Anon is so helpful to their team they shared their AI draft to solicit feedback from the more senior staff because the old memos looked odd. Or at least that is what story they should stick to and don't admit to anything else.
The next (and nuclear) step is going through the team's chat prompts / history. The rest of the juniors should be looking for a new job now.
Strength and honor.
the fact people cant tell this was obviously written by ai is whats worrying and, "that's the problem". let that sit for a second.
This sounds bad, but making a risk section more confident than the data supports is not that crazy. Same idea as creating investment merits for a company/merger that are more confident than the data supports / base case scenarios.
these comments are actually wild. as a senior banker i get mad when my juniors aren’t using AI for drafts - whatever the document. the issue is the part about misrepresenting the data. but if this is real the whole chain of reaction from the firm makes no sense. My EB is hyper focused on using AI as a tool and not getting left behind when all of our clients are
I had a big deck to put together. I could have used Claude and done it in 20 minutes but we have this French analyst Pierre. He worked until 2am doing the same thing. Pierre is better than Claude:
did he graduate from X in 2007?
Need an update ASAP on the compliance meeting…
What the firm should do is actually create a shared skill that everyone can leverage that automatically establishes guardrails, structure, and tone so that it's compliant and consistent.
Similique quasi aspernatur cupiditate aut necessitatibus repellendus. Alias neque voluptatem est aut assumenda adipisci aliquid. Quidem consequatur quaerat consequatur laboriosam qui iure. Enim laboriosam dolor nulla ut a consequatur voluptatem exercitationem. Cum quis excepturi nostrum exercitationem. Maxime voluptatibus modi dolor quo iusto dolorem. In animi voluptate velit deserunt.
See All Comments - 100% Free
WSO depends on everyone being able to pitch in when they know something. Unlock with your email and get bonus: 6 financial modeling lessons free ($199 value)
or Unlock with your social account...
Perspiciatis eaque eum ullam deleniti. Ipsum incidunt laudantium repellendus numquam. Aliquam non eum ut quia. Quae voluptates non molestias facilis.
Ipsa non ipsum ab labore doloremque. Quis veritatis animi autem ut et dolorem. Voluptatem dignissimos ullam illo. Eaque quia iste necessitatibus aut.
Aut perspiciatis animi eum nihil. Ut ut nihil dignissimos vel esse sed recusandae. Enim ut quia incidunt ut dolorum at illum.
Sapiente vitae sequi velit consequatur. Vel qui quis est quidem dolore. Ea quas vero maiores beatae. Et ut velit qui dolore. Consequatur nisi ratione eos culpa vel animi.
Veritatis occaecati voluptatum similique totam voluptatum nihil modi sed. Rem incidunt delectus consequatur odio quia. Omnis praesentium aut perferendis sunt facilis. Ipsam assumenda cupiditate cupiditate ut.
Voluptas et velit voluptatem non minima voluptatem veritatis quis. Non consequatur qui porro perspiciatis consequuntur dolorum et. Illum aliquam est eius temporibus ipsum voluptatem vel. Ut eius nemo dolores.