Andrew Ang: Should AI Agents Run Your Asset Allocation? | #653

Andrew Ang (0:00)

It sounds small, but you compound that over 30 years, and that results in 36% of your portfolio that just disappears.

Meb Faber (0:08)

Is there ever a chance when the agents are like kind of whispering behind your back where they're like, hey, by the way, if you vote for me, I'll vote for you.

Andrew Ang (0:13)

And if you look at this contrasted to like human teams, you know, when you have disagreements, sometimes in extreme cases, HR gets involved or a really talented member, you know, leaves the firm. It's sort of hard to harness this sort of disagreement and differences of opinion, but you can really do it easily with agents, because they're agents, they kind of don't care. Ends of this, the top guy or the top committee and the entering analysts, they're going to be the same. And largely everyone in the middle will have to adapt to do something new or do something different from what they were doing before. The great test, is it better than what we have now for human-based approaches? And to be honest, I don't know. The highest marginal rate for someone in New York City is 55.576%.

I just happened to know that coincidentally.

Meb Faber (1:05)

There's like one investor, used to be two, but one of them passed away that I think really understand this, and it's Warren Buffett.

Andrew Ang (1:12)

Yeah, he's an exemplar, right? Without paying a dividend. If you do the math, 10% in private equity, it's roughly about 7% in equity terms. Or you can do it the other way. You know, once the return in private credit that's needed to match your, you know, 10% in equity, you need like 14, 15% yield there.

Meb Faber (1:31)

It's vastly easier, in my mind, to get alpha from taxes than it is from investments.

Andrew Ang (1:38)

Can't agree more. 200% agree with what you just said.

Meb Faber (1:42)

What's up, y'all? Today's guest is Andrew Ang, founder of Tau Balance, professor at Columbia University, previously ran Factor Investing at BlackRock, wrote a widely used grad finance text. Today he's building AI systems that may replace the human investment community. Andrew, welcome to the show.

Andrew Ang (2:00)

Hi, thank you. Thanks for having me.

Meb Faber (2:02)

You argue that a swarm of AI agents can be broader, deeper, faster than any human investment team. Are we there yet? Are we three years out? Where do we stand?

Andrew Ang (2:13)

I think everything's just beginning. They're being more increasingly used, and who knows what the future will bring, but I think it's going to be completely different from today.

Meb Faber (2:24)

I asked a bunch of people on Twitter the other day, and I was trying to be kind, but I also riled them up a little bit. So I said, because everyone's above average, we know that. So I said, is AI as good as the average CPA today, as good as the average CFP, Financial Planner? How does your swarm look? Is it as good as the average committee or better?

Andrew Ang (2:46)

I think the relevant thing here is the average is going to increase. And I think the right-hand tail will increase out even more.

And so it's not just about whether it's better than an average human. I think it's how much better the average human will get using agents and AI.

Meb Faber (3:09)

Let's get started with this self-driving portfolio. What actually happens from this macro-regime call to a final portfolio? How does one agent push back? Give us the overview.

Andrew Ang (3:21)

I think the agents, they give you three things. They give you scale that you didn't have before. They interact with each other, sometimes for nefarious ways. But I think it helps you to make better investment decisions and gives you new information from that interaction. And then the third one is they can learn.

So, with scale, we took asset allocation process. And there's a couple of ingredients for that. You need some expected returns, you need to formulate a portfolio, right? And you need to kind of explain it, write the investment memos, hold the investment committee meeting, and then execute it. So, the first thing that you can do is really specialize with these agents. Before, you had to have a team. That team set some capital market assumptions, expected returns of like equities and bonds, and some commodities, some alternatives.

And it's usually the same team that does all of that. But you can have an agent now specializing in equities, which is quite different than setting an expected return in treasury markets or corporate bond markets, international equities, small cap equities, right? They're all very different and you get an agent to specialize in each of those. And they can run in parallel and formulate these expected returns very quickly. And it comes with narrative, with plain English explanations and rationales. Yes, they're going to use quantitative methods. These are like the tools that they call or skills that they can use. But you can get scale. And another way is most asset allocation methods rely on just one portfolio construction method. It's usually some variant of old-fashioned mean variants. Thank you, Mark Whitson, 1952 But you can now have an agent for each portfolio construction method. Some of them are pretty trivial, like marker cap weight or equal weight. Some of them are really sophisticated, like hierarchical risk parity or total portfolio approaches.