I’ll preface this by saying that I am not an AI evangelist. In fact, I am probably more a detractor who believes there’s way too much hype around AI. I’ll make a few observations:
- I can see how the technology can be helpful in replacing google searches and maybe even to write a “deep research” note that may or may not be filled with hallucinations. I can see how it’s useful in coding and software engineering but I think it’s usefulness really does stop there, at the surface level. The idea that multiple “agents” can run with a full buyside or sellside process is just plain ridiculous.
- It seems to me that the it’s the most senior people, the talking heads, who keeps harping on how AI will change the world in the next year. These are incidentally the same people who would never have a reason to full test out or use AI models. When I talk to Analysts, Associates or VPs at PE firms, I get a far more tempered response. “yeah, I guess it’s useful to run a quick screen on a private company with public info.”. “Hebbia is not bad but I can’t see it ever replacing our own financial modeling process”
- MSFT AI Chief says most white collar jobs will be automated in the next 12-18 months. I am willing to bet my entire net worth that AI will not make a dent in unemployment rates in the next 18 months. I just wish someone actually hold these people accountable for spewing bullshit like this.
- There are a LOT of people with vested interests with trillions already spent and that will be spent in the ensuing years who need to ensure that this experiment will be a success. They need people who are lukewarm about AI to use it, all day every day all the time. Most of the world’s population just don’t give a shit about this.
- I think the negatives with AI around deepfakes, potential increase in scams, proliferation of AI slop etc will far outweigh any positives. I guess for positives we have, what? improved company margins, I guess. I am not holding my breath on AI being able to solve world hunger or cancer.
With all that being said, I am never one to say that only my view is the correct one. So, AI lovers on WSO, please tell me why and where I am wrong. And in return, I will likely push back respectfully. Hoping that it can be a good discussion on this technology and where it goes from here.
I use AI every day for coding (quant type of role) and it's scary how good it's getting, but at the same time it doesn't seem to be freeing up any time among tech teams at my work. There is no talk of layoffs, headcount on the tech teams has actually increased, and they work nearly as much if not more than the investment people.
I can't square these two things together. Possible theories are:
So in summary, everyone seems to be using AI and getting value from it but I haven't seen a single job eliminated or person laid off. Maybe that means that AI is benign for the average white collar worker?
I'll admit that coding and quant work is very far away from my area of expertise, which is primarily PE with some exposure to real estate and private credit work. From what I have heard, I have been surprised at how useful AI has been for coding. BUT, I 100% believe that the overreaction everyone is having to tech stocks now believing that all these legacy software companies will be obsolete as people whip up their own vibe coded programs is beyond ludicrous.
On your point #2, I think there is an issue here. I maybe in the minority here but I believe there is a shit ton of value in spending the first 2-3 years of your career doing the grunt work - the modelling, the powerpoint presentations, deep research on companies and sectors using that thing between your ears. That's how you learn. That's how you understand nuance. That's how you know what questions to ask. That's how you slowly, in your head, illuminate the world in whatever career you are trying to build. And finally, that's how you earn the right to be more high level and insightful later in your careers - by knowing the nuts and bolts like the back of your hand. If all of the Analysts in IB or PE use AI to do their models, run their decks, type their notes etc, I believe we will be left with a cadre of people down the line who are unable to think logically for themselves. And that negative far outweighs any efficiency gains.
In your view, are companies rewarding or punishing juniors who (ab)use AI too much?
You could make a case for either (punishing kids who send obvious AI slop to their bosses, or promoting kids who use AI to get their shit done fast without checking). At my work it's obviously the latter - investment professionals at my firm do not care if it's done perfect, only that it's done, and investment professionals, not coders, are generally the ones with the power and money at my firm. So it's a no brainer to use AI to churn out as much as you can.
Thankfully, I had a few years to go deep in the weeds before AI did all the dirty work, and I still try my best to learn what I can.
AI "sucks" in legacy projects where you have undocumented context and decisions scattered about within the people working there
At the same time, it's also the case that these newer AI native companies are achieving much higher revenue per employee than any other entity in the past
That's another disadvantage for the incumbents. The current of way of working will last as long as AI native entrants don't replace the large companies we have today
Look at what the finance people are doing at Anthropic. The coding workflow also looks radically different than what exists at large companies
But what exactly is an AI native company? Every company on the planet says it's embedding AI in their workflows and that they want to be AI-native. Can you give me a few examples of truly AI-native businesses that are not selling some kind of AI software?
Think of how much human intellectual cost is buried into many services. OP, you work in AM which is a good example. About 100% of your cost is either direct human labor for your firm, or indirect payments to outside firms (research, legal, banking etc.) that are also mostly educated human labor.
Our most out-of-control industry in terms of weighing on our economy is healthcare. How much of healthcare cost is rooted in the idea that we need to pay mightily for the skills and analytical capabilities of doctors, other care managers and the technical capabilities within the companies that develop drugs and medical devices?
A huge amount of the dollars spent by everyday people ultimately goes to rent-seeking by professionals. Not to mention the insane cost of the education needed to become said professional.
That's the bull case for AI. Life can get a lot cheaper. That's a benefit to everyone. Now for the ~20% of Americans who make outsized income as knowledge workers, it comes at a heavy cost. It's not great for you, me or anyone else reading this. But for most it's a net positive, where life gets much cheaper and the concurrent loss of income is less than that.
I'd love to be wrong. I was rather enjoying my perch in the top few % of earners, and the leg up on life it provides. But I fear I'm right.
If someone wants to help me sleep better, tear my argument apart please.
Yeah, this is an interesting thought experiment. Vinod Khosla went on a podcast a while back and painted a picture of the world like the one you mention. Where you have billions of AI agents doing things that would today require humans to interact with other humans (and this incur a cost). Theoretically, that could bring down the cost of human labor to near zero. In that world, work will be optional.
I asked ChatGPT what % of the world are engaged in white collar and blue collar jobs. No concrete answers of course, but ~35% of the world's population was deemed to be white collar with the vast majority of the remaining being blue collar and informal work. We are hearing more and more about how white collar workers will be replaced by AI. Blue collar jobs are not safe either with all the R&D going into humanoid robots.
I guess my question is, in that world, where everything is so much cheaper, how are people who have been displaced and unemployed by AI supposed to pay (the low cost) for stuff? Please don't say Universal Basic Income. What are humans supposed to do in that world? Because history shows us men (and women but men more so) without purpose and structure is a dangerous animal, doesn't it?
Maybe not exactly UBI but some form of most people getting a degree of subsidy on average, rather than the current system which is most people having to count on the value of their labor to exceed the value of their consumption.
I know pure UBI has its issues. But what about negative income tax, where everyone below a certain income level gets a subsidy (sliding scale) and then above that level you start paying tax.
Ultimately we're going to need a lot less man-hours to generate the same amount of goods & services we generate today. Two knock-on effects will be: (i) higher corporate profits and (ii) very high pay for the small # of man-hours that continue to be necessary under AI. So those now ultra-well-off companies and humans are going to have to have their wealth re-distributed in a way that's meaningfully more than what we currently see in the developed world. Whether that happens through UBI or something more nuanced, I'm less sure. I agree with you that nuance is desirable and a pure UBI system sounds kinda dark.
One mitigant could be, maybe those high-skill humans won't be willing to work more than like 10 hours a week because the extra wealth may not deliver a lot for them in a society where the average schlub already has a pretty good life. That could keep the workforce relatively sizeable, with everyone working less than a full week.
Even if AI automates half of the jobs, people will have way more money to spend because of the savings, and thus consumption will rise.
I generally agree with you -- even in bear case of the avg company being smaller, the bull case that history time & time again shows is then the # of companies increases. Basically incremental volumes will drive growth and opportunity and new job creation, with white collar workers generally being quite adaptable as a category
Moreover, I too find it incredibly frustrating that these AI folks like the MSFT AI chief spew this nonsense with no accountability. This is plain fear-mongering to generate more investors to 1) invest behind a large theoretical TAM and 2) want to own a piece of the future (in part due to fear of losing their own jobs and livelihoods). Anthropic CEO and OpenAI CEOs are notorious for doing this. Even the revs materialize even 1yr later than expected the AI bubble will burst
AI tools are useful in my work no doubt, but it is not a replacement for a buyside analyst with 5+yrs of experience (certainly not for someone with 10+yrs). It is proven in fundamental analysis that more information does NOT equal a better outcome (I've banged my head against that wall enough times) -- it's about 1) finding the right information (vs. table stakes info that isn't going to drive a double in the stock) and 2) analyzing that information to come to the correct decision. Both of which are not made significantly easier by AI
It’s no secret that AI can go off on a tangent, and its systems are only as good as how it is programmed.
Experience trumps everything - the way I see AI being used is simply a tool that can provide information to the users to help make a better educated decision.
I work in tech, and impressed with Claude so far. However, I feel human interaction problem solving are quite insightful since people from all backgrounds can come up with a solution AI couldn’t.
I agree. And with the proliferation of AI slop, I think critical thinking in the future will increasingly mean understanding how to separate insight from data, signal from noise. Because noise and data, I think are about to explode. All these news articles on authors spinning out AI written books in 45 minutes, actors licensing their voice to AI models, everything happening in Sora. I think it's just a sad state of affairs. Creation of the cheapest product or art, sure, but with absolutely no quality or soul.
100% disagree.
AI continues to reshape workflows. I'm surprised it hasn't already for you. Any company research or board briefing I do now starts in Abacus and Perplexity. Any presentation I make starts in Gamma. It's not that AI will automate away my job, but it's increased my productivity by 2x since it has automated the draft portion of anything I work on. Every week I find new ways of accelerating what I do. My job isn't going away, but I stopped using outsourced remote assistants, and I don't need to rely on junior talent. And have you tried vibe coding? I don't build websites for a living, but if I did, I would be panicked, as Replit can build a site faster than one can even dream it up.
Not only has AI not radically changed workflows for anyone at my ~3000 person firm, I know for a fact that it has not done so for my friends working at large banks, many private equity firms and several F500 companies.
I don't mean to throw shade at you but you seem to be a power user of AI tools. Can you tell me more? Do you work in direct PE? What level of seniority are you? Can you tell me how you Gamma, for example?
I work at a REPE fund. Here's what I've done this week in about 3 prompts each:
Of course, I had to review all, check calculations and references to contract clauses, etc., but the initial analysis and review which an analyst on our team might take a few hours per task, was done in 3 prompts per task (about 5 mins of my time).
In terms of impact on staff... I still worked the same hours this week I would otherwise because there's always more work to do. At least the other work was at a higher level.
One final point. Radiologists have been using AI to review cancer scans for years, it is more accurate than a radiologist. There was a study where before and after 3 years of using AI to review scans, the radiologists lost the ability to do the work themselves. This is kind of like some of the more senior people in my firm (esp. origination side) losing their skills in excel.
From my experience: more efficiency at certain tasks, but there's always more work. If you don't check and don't keep your underlying skills at doing the work you automate, you'll lose the ability to do (and possibly check) the work, then AI/automation has worked against you.. us and AI are co-dependent check and balance.
I work for an investment holding company that invests directly in both private and public companies. I'm a portfolio manager (senior investment director level). Lately I've also been interviewing for outside roles. When I have more time I'll add more here.
The point about 'AI slop' is the most underrated part of this. We're prioritizing marginal gains in company margins while ignoring the fact that the internet is being flooded with low-quality, hallucinated content.
Hard to call it a net positive when the signal-to-noise ratio is tanking.
I love to say AI is like your drunk friend at the bar, they’re right 80% you’re just not sure what 80%.
In both Chat GPT and Gemini I’ve had errors recently such as including restaurants in an itinerary that are closed for the specific meal time or ignoring guardrails in prompts like do not repeat a restaurant.
Damn, yeah, I agree that was hilarious and kinda catchy. I might listen to it again. What's your view on human generated art going forward? Is there any point spending years of your life training to be a piano player for example? Any point in trying your hand at painting? Acting?
But I disagree vehemently with one thing you said, and I keep hearing this over and over again from the tech overlords. "Upskill or be obsolete". What exactly does that mean? Upskill in what? If I am a rapper, how or in what am I supposed to upskill when AI can now make a song like the one you have here? One that I imagine was "inspired" by countless other hip hop songs created by humans.
What song?
There is no chance you can argue that song isn't AI Slop.
Stop it.
Throughout this entire thread, I was trying to keep my mind (very very) open and give him the benefit of doubt. Not easy with all the AI lovers around.
The point about 'AI slop' and deepfakes is often ignored in favor of margin talk. Until these models can handle complex financial modeling without needing a human to fix the errors, the hype will remain just that, hype.
Am I the only one who feels like AI companies don't focus on the right areas? It can't make practice questions for my AP classes or my SAT, but it can make a hyper-realistic video of Jeffrey Epstein making a TikTok dance. make it make sense.
I have already slowed down hiring and realized major margin improvements at our portfolio of consumer brands. I don't think people realize how fast it's improving.
Think this is true across many companies. Though my thoughts are on are that jobs needs to be redefined with AI tools in mind and job duties/roles will change drastically over time. I’d wager more jobs will be created in due part thanks to AI, not reduce the job market.
I know you are a legit poster on WSO. So not that I don't believe you but can you give 1 or 2 examples of margin improvements in the companies you are involved in?
I work in private equity funds and co-investments. In our portfolio of the larger consumer focused PE funds, I have seen no or very little margin improvements in the last few years. In fact, Apollo had shown a chart on their linkedin page recently that said the only sector that has seen margin improvement (likely from AI) over the past few years is the software sector. And this is for companies in the S&P 500. Every other sector has the margins today that they have had over the last five years.
OP here. Here's what I think is another fundamental issue with AI. And this one's a bit out there. I can't prove this but I have a feeling that the constraint to more efficiency today just might be the human brain rather than compute or automation. Or at least we are getting pretty close to what the human brain can effectively handle. I am a PE guy, so the deal side is my context here.
See, when you are on a deal sprint, you need to know every last thing about that investment. Sure you may be able to automate the modelling and maybe even some of the lower level analysis to AI but then what? You take on another deal? another 2 deals? How many Principals can keep straight in their head the key numbers, workstream progress, legal & tax materials on three live deals at the same time?
I wonder whether at some point we push the idea of efficiency past the capabilities of the human.
This is a good point - I've used AI in a lot of ways, but you still have to check what it does. I'm handling more now than ever, but even automating 100% of what I do and leaving me with the review only, I'm starting to feel a bit overwhelmed with the mental bandwidth to manage all this. It's not a time issue anymore, but there are still bottlenecks.
Fascinating. I had never thought about this before but I totally agree. The analysts can whip up so much now that there’s way more work because we have to dig into all the details of a million things all at once all the time.
This is a pretty sensible perspective, in my opinion. Those at the top appear to generate more of the hoopla rather than those who perform the work on a daily basis. For the majority of individuals, artificial intelligence (AI) is more of a time-saving assistant than a replacement for full-time employment. Expectations seem to be far higher than what technology is now capable of.
If you think convincing deepfakes and convincing scams will proliferate then that must mean there are serious commercial applications because high quality video/audio/photos and narratives can also be legitimately sold
Eius architecto fuga mollitia aut incidunt inventore non. Ea quam ducimus inventore minima. Est odio assumenda non ea. Omnis quos rerum est ipsam laudantium. Error minus occaecati accusantium ducimus laborum. Aperiam esse cupiditate aut praesentium dicta.
Perspiciatis aut asperiores voluptatem voluptas. Incidunt corrupti cum minima et. Sequi culpa incidunt sed sit corporis. Dicta at sint qui ut temporibus et id.
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...
Voluptatum dolor tenetur nihil. Et neque beatae quo quod ut non voluptas. Veritatis quia ipsam ducimus est quia quo et quae. Accusantium reprehenderit non libero sit cumque.
Necessitatibus provident a iusto pariatur tenetur nam. Libero molestiae molestiae perspiciatis velit laborum. Voluptas rerum aut fugit rerum voluptates dolore. Et nemo quo qui magnam itaque. Et autem cupiditate voluptate tempora.