AI can now draft 95% of an S1 IPO prospectus in minutes
GOLDMAN SACHS CEO DAVID SOLOMON:
"AI can now draft 95% of an S1 IPO prospectus in minutes (a job that used to require a 6-person team multiple weeks). The last 5% now matters because the rest is now a commodity!" THE AI EFFECT AT ITS FINEST
Has AI really advanced that much?
research is being published demonstrating AI's superior reasoning capabilities in the medical field
All traditional industries are going to be disrupted
Can any GS bankers confirm the s-1 process being radically different thanks to AI? “Can” and “Is” are two very different statements.
AI in theory can do every part of our job, with the adequate information being flowed into it, and adequate template provided, and adequate guidance on how to improve it, and adequate review of its output. AI’s ability isn’t debatable, it’s the successful implementation that will get in the way.
On 2 early stage processes but the window isn’t open yet for my sector. Will report back when we’re drafting
Bumping this 1 year later - anybody can report material operational differences yet?
I mean, it's not like all S1s were previously written from scratch. You were always going to work from an existing document / template and then iterate the last little bits to be company-specific. I'm sure AI can accelerate that, but I doubt it's that fundamental of a change.
DSol has an interest in looking tech-forward for prospects / shareholders, and that incentivizes statements like this. It's the same deal as Klarna loudly announcing they were laying off people because of AI and then quietly rehiring them.
Think this is right. It's the last 5% that has always been the hard / value-add part. AI can do 95% because 95% is boilerplate and recycled from the last similar S-1.
He is a bit of a clown, but not to the point of making someone believe that his team was spending 5 weeks in the last decades preparing things that they already had templates for. He means the parts that were actually worked that needed individual output.
And you know this how? He told you?
AI can also write a 10000 word essay in minutes. That doesn't mean you're going to submit it though
I think bankers (for the most part) are remarkably ill-educated about AI. I know high school students that use it (dishonestly) to fake research papers that get published in top-tier journals where the vast majority of PhD-written papers are rejected.
The white collar workforce of junior bankers, consultants, computer scientists, and non-surgeon physicians will be 95% reduced by 2030, 2035 at the latest.
Although your claim that high school students are beating out PhDs to journals seems untrue, lets assume it is for a second. There's a significantly large difference between submitting an AI written paper to random journals to drafting an IPO prospectus for a multi billion dollar company. For starters, there is effectively zero negative repercussions to submitting a paper with errors in, if the errors are caught who cares the journal will just reject it. If the errors aren't caught, it doesn't really matter apart from the academic community finding it unethical.
Now, can you imagine a company spending millions in fees accepting something like that? Imagine a banks reputation if the AI work has errors? This isn't even getting into the monumental regulatory risks here (everything here is SEC regulated). As commenter said below all it takes is AI erroneously writing "will' instead of "may" for a billion dollar lawsuit.
Yes you can use AI to get an IPO prospectus in minutes, but until AI is completely airtight 100% accurate, it will still require days of analyst time to check everything, and effectively do the work themselves anyway.
I don't know how soon we will reach that point, but nearly everytime I use chatgpt I find myself correcting errors it has given me in its responses. And until it does then analysts will be required to do everything themselves because it's much easier to ensure accuracy in work you have done than checking through someone else's.
Public models get 3-5x better (codeforces) every year. Do you really think they will be making mistakes in 2030?
Sorry for the long responses. I’m about to retire and have a lot of time.
Lol, we r still in early stages of AI. this is like being in 1980 and saying personal computers could never work because the screen is too small.
Can you give the source on the high school students doing this? I can see one done by sakana.ai but they were a 4-man team who are working professionals / professor from Oxford
Obviously there is no source, this is academic dishonesty and will get kids kicked out of school and rescinded from college. I know peers of my children that have done this. Power to them I guess.
The whole healthcare thing makes me laugh. What happens if there’s a doctrine of strict liability? How is AI going to be doing physical exams? What about patients who are poor historians? Will AI be getting prescriptive authority?
Lol. Thanks for the laugh.
woah, I might consider going back to law given that now everyone relies on AI to draft and write things, so might as well capitalize this laziness by litigating over damages for misplaced words that some are not even aware got slipped in the doc by AI
Imagine a full S-1 of more than 200 pages being done by AI, which no one will read in the same detail vs. when someone drafts it from 0 carefully, and using "will" instead of "may" so then the company becomes liable to investors for not undertaking whatever action they put "will" in front to hahaha
what a fucking blast, probably the new field for a finance + JD advantage similar to what distressed debt was in the 90s.
seriously considering it.
Exactly lol. If AI takes "minutes" it's still going to take days to proof read everything/do all the calcs yourself to check the numbers
just need the bottom of the pyramid monkey to ctrl f "will" and you will be out of a job...
you put too much faith in our intern in question
I feel like executive management believes AI can do anything. In reality, the people using it day to day understand its limitations.
Yes, AI can write 95% of a shitty S1 which is not a good enough product. As others have said, that barely gets you any further than starting with a template or someone else's S1.
Reeks of boomers who haven't done any grunt work in 30 years but are desperate to be seen as "keeping up with technology"
Yes but you have to understand that AI is at the stage that smartphones were at in 2000. It’s getting 5x better by the year (codeforces).
This is correct. Martin Chavez was CFO of GS (sits on the board of Alphabet) and said "bankers have this hubris about them. Yeah..Sure... Tech is eating the world, but we're special." They're idiots.
I think the AI jump will be like the jump from pre-computer offices and white collar work to getting excel and word on a computer.
I'd prefer he use the AI to put himself out of his DJ job
This is not a critique of the substance of what David said, but how would he actually know? He is so far removed from the actual work. It would take him going through layers of management to get to the actual people doing this work.
This strikes me as a marketing ploy more than an actual understanding of the technology. OR he is being intentionally misleading with how most of an S1 is going to be boilerplate language anyways so an AI likely can pick up that part pretty easily, but he isn't giving this part of the explanation.
I) this is definitely a D-sol marketing ploy. When do you think was the last time he wrote a S-1 or even asked how mechanically one was made
II) I agree that AI can get a S-1 80% of the way there… along with most other workstreams in banking. It’s the last 20% of company-specific stuff that takes the most thought, time, and effort. AI will definitely save some time on the front end putting together a shitty document that needs to be reviewed and reworked, but that’s the easy part.
Tbh what this means is that investment banks will just churn out more work for identical tasks as the expectations rise from clients. Someone still has to coordinate, make assumptions, organize, etc.
Ullam accusamus ex unde. Odio unde sunt placeat. Dicta earum aperiam distinctio id ratione.
Natus laboriosam voluptatem et in. Laudantium consectetur eum fugit. Molestiae ea natus pariatur aut vitae.
Rem vel eos et aspernatur. Sunt est consequatur et modi nostrum laudantium.
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...
Aut rem minus consectetur atque sit. Sint iste corrupti qui autem. Quas dignissimos quis doloribus voluptatibus. Excepturi consectetur ex culpa fugiat.