We Always Used Rules to Take Emotion Out. Then AI Changed the Game | Glenn Caldicott

What we talked about.

Sunny Ray talks with Glenn Caldicott, chief investment officer at Empirical Asset Management, about how rules-based investing has evolved since he wrote his first rules in 1994. Caldicott explains that emotion, specifically fear and greed, drives most investing mistakes, and that rules exist to remove it. He recounts building his first strategy around clusters of insider buying, tested with pen and paper before spreadsheets or the internet were available. The conversation then turns to AI, which Caldicott frames not as a break from rules but as an extension of them, adding analytical capacity within a man and machine partnership. He describes working with an AI firm he calls Treface, whose closed loop, auditable platform builds portfolios that Empirical delivers to clients. He also outlines how the firm is organized, how clients find it, and why it plans to launch three ETFs that invest with AI rather than in AI. Caldicott closes by warning against black box systems where the machine becomes the decision maker, and shares where listeners can learn more.

Glenn Caldicott says AI doesn't replace rules-based investing, it extends it, as long as humans keep control and the system stays transparent.

The questions, and the answers.

What did your first year in the business teach you about risk?

I learned you had better understand risk, and if you don't, stay away from it. Half of what we do in rules-based investing is risk control, not chasing performance. Ironically, when you measure and manage risk you tend to get better performance, and when you chase performance you tend to get worse results.

What went wrong in the early days that made you want rules at all?

I was helping clients build portfolios and I was decent, but only 50/50 at best, and most people were worse. Most managers still get outperformed by their benchmarks. I realized the problem was emotion, fear and greed. Nobody I've ever asked says they never made an emotional investing mistake, so I decided to get emotion out with rules.

What was your first set of rules based on?

I looked at SEC insider filings. Insiders sell for a million reasons, like a boat or a divorce, but they buy for only one reason, to profit. I looked for clusters of buying by people wearing different hats in a company, built rules, and tested them with pen and paper. That became my first rules-based strategy.

You spent decades arguing rules save people from themselves. What changed your mind about machines?

Nothing changed, it's an extension of the same thought. Rules save you from yourself, and AI is just a tool. It's man and machine, not man versus machine. It takes time to get and clean the data, build the algorithms, and keep training the system. All it really adds is capacity, like having thousands of analysts collaborating.

What decision will you refuse to hand to the model?

The end result has to be controlled by the human. The three AI companies we work with use closed loop systems, not black boxes. Everything inside is transparent and auditable, and it doesn't go out on the internet or use large language models. It does numerical calculations at massive scale, and we make the ultimate decisions.

How do the right clients find Empirical today?

Trust builds over time, so we get a lot of referrals. The firm has grown from 100 million to 700 million in about four and a half years. The second path is content from Treface, which embeds links on big sites so investors and advisers can set up a Zoom meeting with us.

What's coming next for the firm?

We plan to launch three ETFs around the beginning of next year, one from each of our AI companies. This is investing with AI, which is different from investing in AI. We're using AI to build the portfolios, and it's a niche we'd like to own to a large degree.

What happens when you try to protect the term rules-based investing?

I trademarked it in the 1990s, assuming everyone would have to do this. For 10 or 15 years, nothing happened. Now everybody uses the language, and defending it was turning into a million dollar a year game. My intellectual property attorney told me to stop. I can still use it, but I can't protect it.