AI Creating More Work
I’ve had several experiences over the past few months where my clueless MD has stopped by my desk, gave me some obscure ask, and said “don’t spend more than 15 minutes on this just use AI!”
Meanwhile the task will be some ridiculous ask that AI cannot do, like putting together a 10 page, client friendly industry overview deck.
While going through a CIP drafting session, my MD constantly goes “oh let’s add a page on ___, AI should be able to pull it together in a few seconds.” After 10 more new page creates that “the AI can do”, I’m up till 3am putting pages together by hand.
Obviously AI can do some pretty incredible stuff, and every day I’m finding new ways to save time with it, but as people start worrying about shrinking analyst classes and mass layoffs, I wonder if instead the standards for materials and overall workload will increase to counter this.
Who knows how this will all play out, maybe Claude will take my job, but wanted to share a perspective.
I'm with you. Experienced something quite similar recently.
I did push back that I'm pretty sure what he was asking for was not going to be an AI one shot but he didn't fully believe me.
It will only reduce work and cause layoffs if ONLY juniors had access to said tools. The amount of work that is now created with MDs abusing AI more than enough offset the productivity boost. They now
Asking AI for what "the deck is missing" which, shocker, the AI agent will always suggest more and more pages with nonsensical guidance
These retarded boomers legit think AI is a magic answer box. The amount of times I get emails telling me to "run the deck through AI" to see what else we should add is maddening.
This
The best part is that this is all going to culminate into a bunch of firms with undifferentiated advice and materials leaving a clear gap for people who can still do their own thinking. I doubt owners and C-suites are going to select the bankers who put AI slop in front of them in the next 5 years. They might be able to get away with it for now given the potential gap in knowledge in AI, but people will catch on.
Additionally, I am a believer that AI will become more expensive and when that happens are, people really going to pay $10,000-$20,000 for a subpar CIM or pitch. Maybe, but I find it unlikely.
Bro they be thinking AI is god, yet my firms shit ass copilot can't even organize a list alphabetically
It’s so easy to notice that the deck was created using AI. You’ll see a bunch of nonsense crammed into one slide and MDs think they done something
AI has raised senior expectations faster than the AI capabilities juniors get
Just to add to this, I’ve also noticed much lower quality work being pushed out by seniors. Some senior bankers I work with now will send a full AI deck or write up to a client. Filled end to end with total nonsense, errors, hallucinations. It’s been shocking how quickly they have just accepted low quality work.. clients obviously know it’s AI, just pure laziness.
Would’ve never happened six months ago.. would have been fact checked and reviewed by multiple people.
Might just be my bank? Anyone else?
EB or BB?
ROTH/GUGG/PWP
Good lord. Really??? I would think the clients wouldn’t be too excited about paying seven figure fees for “advice” they could get for free from claude.
The worst is when a MD uses AI to question your work, and when you look at his output its just hallucinated BS.
AI output is only as good as the prompters input.
AI has made this job way less enjoyable. I think there's a lot to like about AI in IB workflows (stuff like categorizing thousands of rows of data in Excel or mapping two different database outputs together). But a good 40% of my job now is inputting something to AI -> getting slop output -> fixing AI slop output. And fixing AI slop is basically fighting the AI going back and forth 20x to get to something workable, then manually editing text / output onto PPT.
At this point even tho I work in TMT I'm hoping AI just completely flops and it gets delegated to a helpful chatbot that can help increase productivity by 10%.
amen. Job has gotten materially less interesting in the last 6 months
Fixing AI slop is arguably more manual and more painful than everything AI automated. But then you are forced to start off with AI slop because MDs are now allergic to any real guidance or even real comments at this point. A lot of time they just send AI slop to us as "outlines" to "refine".
This hits home. My team just wrapped up a deal where >70% of the outputs were almost entirely created by AI, then formatted a little bit
well, if your MD said “don’t spend more than 15 minutes on this just use AI!”, then just type whatever he said into the AI tool you use and send it to MD. if he doesn't like it, remind him that he said “don’t spend more than 15 minutes on this just use AI!”.
try this and see how quickly they scream at you like a banshee lmao
AI should be able to do a 10 page industry overview deck at 80%+. Do you have precedents of that industry and a connector that accesses all your precedents? What tools do you use?
Agree lots of MDs overestimate output without proper investment in databases and upfront setup, and slop in = slop out. But good prompt + good precedents = 80%+ on exactly that type of example.
Which areas does AI fail the most?
For most of my career, I’ve looked at the quality, depth, creativity, and attention to detail of a potential partner, vendor, or employee’s work as the IQ test.
The Principal-Agent Problem (can I rely on these individuals) and finding out if guys were just beneficiaries of a good market and timing or have something special.
That’s how I approach first time deliverables to new capital partners:
“The assignment is an IQ test of our firm. They already know the answer and are withholding key information, and want to know how resourceful we are and whether we would ask the right questions to get to the heart of the matter.”
It’s become harder to gauge with AI used more and more especially at the vendor level (since I’m seeking out outside expertise), but also more at the partner level. It’s not that I don’t want them to use the tools (especially if they are older and run very lean without analysts), it is that it’s harder to evaluate what they already know and how they think on their own.
At the end of the day, the results matter. So, presentation ability becomes less of a barrier to entry; not a disqualifier but also not a gauge. Track record becomes more important. And probably other factors I should be relying on more, such as asking good questions, however, I believe AI will help you with that also.
Would you say this means the AI models are approaching the high IQ work you've seen in the past (in terms of quality, depth, creativity, and attention to detail)?
Yes and no. I can tell it’s AI, so it just checks the box of the answer might be correct, but I don’t have any idea if the person thought of the answer themselves, which is what I was really after.
They genuinely think that the AI tools the banks have are as capable as Claude. Rogo (which many banks use) straight up makes up numbers all the time.
Claude does the same shit
Is it actually any good for anything that OpenAI/Anthropic cant do
It's worse on the corporate side. Non-technical people are empowered and any idea they have can be turned into a professional and well-thought-out-appearing analysis. You try and validate some aspect of their thesis and you see thousands and thousands of formulas like this.
=IF((COUNTIFS(Externallylinkedfile!$A$5:$A$204,TRUE,Externallylinkedfile!$B$5:$B$204,"LIGMA",Externallylinkedfile!$C$5:$C$204,$A15,Externallylinkedfile!$D$5:$D$204,N$11)+COUNTIFS(Externallylinkedfile!$A$5:$A$204,TRUE,Externallylinkedfile!$B$5:$B$204,"LIGMA",Externallylinkedfile!$C$5:$C$204,$A15,Externallylinkedfile!$D$5:$D$204,"All")+COUNTIFS(Externallylinkedfile3!$A$5:$A$204,TRUE,Externallylinkedfile!$B$5:$B$204,"Deeznuts",Externallylinkedfile!$C$5:$C$204,$B15,Externallylinkedfile!$D$5:$D$204,N$11)+COUNTIFS(Externallylinkedfile!$A$5:$A$204,TRUE,Externallylinkedfile23!$B$5:$B$204,"BOFADEEZ",Externallylinkedfile!$C$5:$C$204,$B15,Externallylinkedfile3!$D$5:$D$204,"All"))>0,SUM($G15:$I15),0)
They can't explain what's going on and especially can't explain why the numbers change every 20 seconds.
Ha gott'em!
Trying to get any AI to do things the right way is an incredibly frustrating and borderline adversarial exercise. No matter how specific you get, no matter how many examples you provide, it will fuck up. It takes more time and brain power to engage with AI's shit outputs than it is to just do it most of the ask yourself. Right down to things that seem like they should be simple for a LARGE LANGUAGE MODEL. For example, you Mr. GPT have access to all of my emails, messages, previous work etc. Why the fuck cant you write in my style of writing, sentence structure, verbiage, word choice, etc? No, it can't do it. Instead if gives some pseudo liberal arts slop taking 200 words to say something that could be said in 20. It has no ability to understand what is and is not important, it makes shit up, it completely misses obvious things. Sometimes I'll just paste its output back in without reading and ask "why are there errors in this?" and it will respond with "Good Catch! It seems I inadvertently did X Y Z instead of A B C". FML.
Realatable, it pisses me off as well when I read things like the bolt part. It's like reading "yeah, sorry, I knew it was wrong so I just made it up hoping you wouldn't notice!!
Just wait until the client starts raving about AI. Even more painful
It's fun to sit with other people when they prompt AI. You'll know by the first prompt if they're retarded or not. What usually happens is they are indeed retarded but are the biggest braggarts about using it. Ignorance is bliss, as they say.
lol, it really is a great litmus test
Had some terrible experiences with a boomer MD who thought AI could do everything. When I'd give him work to review that was done using AI per his instructions, he'd say that this wasn't what he wanted. I even used the same prompts he was using and cross-checking against my more refined prompts, and he would still be unhappy. Would rather dip my head in acid than have an MD say, 'oh don't worry, just use AI'
This was helpful 100 % no cap
Agree with everyone here but I will present a more cynical angle that I don’t think most people are focusing on: MDs largely know AI won’t cut it and that whatever they are asking will blow up your night/weekend, but they say “AI should do it in 15 minutes” as a way to gaslight you and remove personal liability for when you do pull an all nighter doing their last minute ask. This has happened to me time and again. They are just trying to pretend like what they are asking you is simple, so that they are not the bad guy. Many firms now also chase down MDs when juniors work all nighters or are logged on for over a certain number of hours, and this is their new alibi.
Agree with the observation. But this has alreadybeen prevalent way before AI. Always the same schtick: asking for something that takes at minium 2-3h if done properly and then prephrasing it "lets just add XYZ, anyway should not take you more than 30 minutes".
It is done v intentionally because it shifts the blame to you as they position themselves as having given you an easy task and if you end up working late for that is somehow your fault that you made it more complicated as it should be.
sounds like a UBS ED
bump
Voluptatem est molestiae soluta ea aut quaerat delectus. Qui accusantium quis nesciunt quasi.
In consequatur eligendi perspiciatis accusamus ea eius aut. Ut in debitis eos accusantium illo laudantium quas. Modi nostrum eaque cupiditate. Sequi nihil autem perferendis labore accusantium.
Illo aut nam in voluptatibus vitae cumque aut. Repellendus atque nostrum et dignissimos laboriosam ratione ut. Temporibus praesentium quia error voluptas quis autem. Saepe beatae numquam explicabo aliquam amet voluptatem. Et dicta ut ipsum dolor eligendi iusto sapiente.
Unde qui repellendus ipsam voluptatum voluptate. Qui fugiat nemo deserunt. Voluptatum sit officia aliquid quaerat corporis ab dolor. Ipsum ut et error tempora. Eligendi molestias reprehenderit rerum eaque ut fuga quisquam.
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...
Neque quo earum ut vel. Sit et quasi nemo. Officiis nihil sint explicabo et fugit asperiores magnam distinctio. Voluptatem ut voluptas et pariatur quisquam amet maxime.
Consequatur aliquam non voluptatem iusto iste nisi quo. Explicabo non incidunt veritatis aut. Hic voluptatum dolore ut laboriosam dolorum. Autem ut pariatur aspernatur nihil.
Et mollitia animi assumenda aut qui aut. Numquam velit qui debitis et. Illo occaecati et dolor pariatur est hic labore. Distinctio facere et ullam quisquam.
Ducimus temporibus sit id sequi eos minima doloribus. Perferendis laudantium nesciunt quae et maxime et corrupti.
Dolores sed numquam commodi. Accusantium rerum facere ea officia vero iusto est modi. Consequuntur magni illo similique illum laudantium.