# Faculty Talk: AI in the Boardroom

## Метаданные

- **Канал:** Columbia Business School
- **YouTube:** https://www.youtube.com/watch?v=pTtOd9fEHdo
- **Дата:** 21.07.2026
- **Длительность:** 11:48
- **Просмотры:** 98

## Описание

Shiva Rajgopal, Roy Bernard Kester and T.W. Byrnes Professor of Accounting and Auditing; Chair of the Accounting Division.

## Содержание

### [0:00](https://www.youtube.com/watch?v=pTtOd9fEHdo) Segment 1 (00:00 - 05:00)

Um two things before we start. I'm very happy AI is boring. Where's Sandra? Uh oh. And then uh I'm just here to sell books, okay? So, buy early, buy often. Uh so So, let me give you some background behind what this is about. So, I'm an accounting professor, and we like boring. Uh And I spend a lot of time with uh you know, educating people who are CFOs and people in boardrooms. And you know, over time I've begun to realize that being on a board is an impossible job because you rely on management to give you information to evaluate management. And that never works by and large. I mean, you know that. Many of you are on boards. Uh people often complain board members are overpaid. No, they are not because you're barely paid for six meetings, three days per meeting, travel time. And if you're board worthy, your wage rate is at least a thousand dollars an hour. Do the math. You're probably underpaid. And what's the point here? To a great extent, you know, it's an information problem because management is immersed in what it does. Board members show up for maybe, you know, three days, five to six times in a year. Management has a whole bunch of consultants working for them. But the board doesn't. Very often boards don't have a separate consulting budget. So, this is where AI comes in. You know, whenever I used to teach these classes, I'd get all kinds of questions. Can you write me a board resume? Who am you know, what kind of industry am I a good fit for? I have an interview tomorrow. Can you go through the 10K and help me figure out what is this business about? Can you read the proxy statement? Can you Can you Can you get me a sense for what the other board members are like? Who's the person there trying to protect the CEO? who's a cost cutter? Who's a revenue growth mindset kind of person? You know, and it's Okay, it's very expensive to do this on a one-on-one basis, but I do it because I'm a nice guy. But now you don't have to do that because you have AI. AI is so incredible, in my opinion. I'm so happy with the Baskin-Robbins and all that because a lot of what I used to do, people would sit in class and would appreciate that. It was impossible for them to scale that. So now I can put out my model of how I think. It may not be the best model, but it is a model and it'll give you reasonably good answers. So that's the whole project. So uh — [clears throat] — Uh Yeah, and uh you know, we can talk about all these uh postmortems, which we don't have to go into. But the larger point, and I've been on boards, and many of you are on boards, by and large, groupthink is a huge problem. People know there's something wrong. They either don't have the vocabulary to express it. They want to be loved and respected, hence they don't want to be seen as a painful person. Some of them are on boards and they actually value the money that they get and they don't want to get fired, right? And I don't know how to solve all that, okay? I'm just giving you a technical problem. If you don't have the will to govern, then I don't know how to solve that issue. But at least now you have the tools to actually go and do a bunch of stuff yourself. So this is this a slide that's meant to scare you. You're supposed to have legal liability, but in reality, you have D& O insurance, you can get lawyered up, and you can you can't get sued. So forget this for a bit. So let's let's just assume that you actually want to govern and ask questions without being a pain in the butt, but just because that's your job, right? Without AI, what would you get? You'd get a board book, 100 pages, two, three days to review. You kind of skim something. Maybe you stick it into ChatGPT. You get some answers. You ask some questions. You have a very nice dinner and you go home, right? That's a standard board meeting. Well, hopefully with what I'm going to give you, you can do much much more, you know? You can actually look through a proxy, look through a 10K, look through everything effectively, you know, earnings transcripts, conference calls, stuff that the hedge funds and the short sellers can do now with AI. It's kind of remarkable to me that people inside the company who are supposed to be responsible for overseeing the company don't have access to. That's basically what I'm trying to solve, you know? Um Examples, and I have like hundreds of pages of prompts behind this. I just picked one company that most people would be able to relate to, Netflix. So, what's the biggest What is the biggest accounting issue in Netflix? Question simple. What do you think is the economic life of a Netflix show? Q& A was not allowed. I just get bored talking to myself, but What's the economic life of a Netflix show? Just give me some numbers. Nobody thinks 1 week? I think that's what it is. You What is What is Netflix's business model? It's to put out a each show every week. Why? If I don't like you, I'll leave you, right? So, you know, this is all very interesting. So, what kind of

### [5:00](https://www.youtube.com/watch?v=pTtOd9fEHdo&t=300s) Segment 2 (05:00 - 10:00)

accounting useful life does Netflix assume? 4 years. And that's okay. Maybe it's in a portfolio, who knows? I don't see their viewing patterns and so on. But, what's interesting though is that the entire balance sheet falls apart if you relax that assumption, right? Obvious point. Why? The biggest asset are content assets, right? If you're going to depreciate this 1/4 at a time, you can visualize this, go back to your horrible accounting classes that you don't want to recall, but, you know, well, if it's 1 year, the asset's all gone. And where is that going to come from? From owner's equity. Basically, you have no owner's equity left. Right? So, just to be able to run simple thought experiments like this used to be difficult if you're not a finance person, right? And if you can dig deeper, Netflix gets two types of content. You know, they license content and they produce their own. Right? The economics of these things are hopefully quite different. Which of these things do you would What's the riskier way to acquire content? Is this Is it a license or is it to make your own? — Make. — To make your own. Because to make your own is like the oil business. Right? Like nine dry wells will pay for the one wet well. But Netflix's economic life, at least in the accounting financial statements, for both licensed and produced content is the same. Which should worry you if you're a director. How can they possibly be the same? So, things like this, okay? This is just a simple case. We can go on and talk about other things. But it virtually It works very well for most companies. You know, the S-1 for SpaceX was out and I could do something in like 4 hours and I put something out. I can I can't read the S-1 in 4 hours, okay? For a SpaceX. It's like what? 890 pages or something like that. So, the idea is simply to give you tools in a more structured way. You can ask me, "Why can't I just stick this into a 10-K and ask GPT to give you something? " My worry with AI is that it is so elastic that you can do anything with it and virtually every statement about AI is true. A lot depends on what you ask. how you sequence the prompts. And this is hopefully the value add from this whole experiment. Now, let's say you're on the comp committee. Right? Um So, it turns out that Netflix has two CEOs after Reed Hastings left. Uh they're co-CEOs and they're both paid roughly $60 million. Which is a lot, right? So, thought experiment. It seems like a lot, the wow, how much lot? So, I can basically have my prompts go and say, "If Netflix were a standard S& P 500 company, how would you pay these people? Keeping everything else constant, okay? ROE, profitability, whatever else you want to keep constant. " It's barely 13 mil. So, 47 mil must be for something. Talent, twice the talent to replace Reed, and it turns out it's all because of benchmarking. Everybody in the media industry is massively overpaid for whatever reason. You know, most companies in the media industry barely earn cost of capital, right? So, at least it gives you some data points to go ask some questions. Assuming you want to ask them. If you don't want to ask them, as I said, you I don't know how to help you. But, at least there's some evidence-based, data-based discussion that's now possible. And, by and large all this can be done in like 15-20 minutes, right? The hard part, of course, is you have to make sense of it, for which you need to know what you're doing. You need to have domain expertise, all the usual stuff that I'm sure people tell you about AI. So, so what am I doing? I'm just taking the whole value chain. Are you board ready? Do you want me to write you a resume? tell you what you're a good fit for? How does the interview work? I would usually get, you know, some of the recruiting firms to come and interview people in my class. They'd usually pick one person. They don't know the person. It's a nice mock experiment, but what about the other 30 people? Here, you can do all that. You can customize it to your resume. You can set the temperature. And, aggressive person who's only trying to protect the CEO's job. Somebody else who's an institutionalist. Somebody else who cares about rules. What kind of questions will you get? You know, the potential is infinite. That's why I'm so happy to be a part of Boring AI. And, then you can go down the chain, right? I have an interview in about 2 days to be on a board. Can you go through the 10K proxy, everything that's going? And, at least and help me understand how the business makes money. How do materials, labor, capacity, and talent combine to add value? If you're on the audit committee, what should I worry about? If you're on the comp Allocating capital. Now, capital actually has a cost. Earlier, we had 0% interest rates. How do you allocate capital? How do you figure out whether you should do buybacks or R& D, CapEx or R& D? Basic stuff like that, you know. Scenario planning. You know, as you know, I mean in the last 15 months, risks in the world have gone up. So, are you prepared? Let's say the straight of almost shuts down. What do you do? Again, with some reasonable discipline and with AI, you can do a lot of scenario planning. Sustainability, I put that at the end.

### [10:00](https://www.youtube.com/watch?v=pTtOd9fEHdo&t=600s) Segment 3 (10:00 - 11:00)

It's a bad word these days in this country, but you know, it's a big deal. And there's a lot to be done on sustainability as well. So, but again, sustainability asks an even harder measurement question. What value does this company add to society? How do you put a price on consumer surplus? say other externalities? That's what we tried to do in that book. The usual worries, oh, doesn't AI make up stuff? You know, not in my experience. I think the models have become so good. Maybe 1% of the time, maybe 3 4% of the time. Of course, you have to pay attention. Right? The last two or three percent cannot be commoditized. The other stuff, you can get the machine to do. Confidentiality, yeah, general counsel is going to be the biggest problem in adopting any of this, right? Because most lawyers in general, sadly in my experience, haven't taken a technology class. But they they will be the last mile before they let you do anything. So, that's why I'm saying do this with public data. There's no confidentiality issue there. Liability, that's always an issue. I guess you just go by D& O insurance. Uh So, bottom line, you know, stuff like the AI makes up stuff and things like that to me are not such a big barrier if you really want to do this. So, bottom line at the end of the day, I have 1 minute. The hopefully in 10 years or 5 years, the question will not be did you read the board book? You know, it would be did you do some due diligence? Do you actually want to govern? And you know, I came here to do a job. Now give me the tools. So, you know, take some pictures, buy some books. —

---
*Источник: https://ekstraktznaniy.ru/video/53389*