You can’t spell liar without AI

Jack Meskunas, of Oppenheimer & Co. takes a look at why AI can come with a catch.

I’ve always valued the energy of captive insurance conferences – the travel, meeting new people as well as seeing old friends and most importantly the inspiration for new articles.

Is it expertise, or is it AI?

I had a fascinating conversation with an insurance executive who lamented that it has become next to impossible to determine what and how much someone knows about technical details and specifics in the insurance and reinsurance industry. Why? Because job applicants have been able to highly customise their CVs - even to the point of submitting white papers on the job responsibilities they expect to have - and AI can write these for them, making them look like in-depth subject-matter experts.

The catch? Once you get to the Zoom or in-person interview, the illusion fades and you’re left with someone more adept at prompting ChatGPT than navigating insurance structures.

Some might shrug this off as no big deal. They got caught and didn’t get hired, so no harm done. I disagree. This kind of behaviour ultimately costs the hiring company time and money in addition to lengthening the time to hire. It also sheds a very cold harsh light on an applicant who would so blatantly falsify documentation in a desire to get hired into a position that actually requires experience, knowledge and skills.

In insurance, that could mean lost money; in other fields, it could mean lost lives. Dishonesty of any kind - and that includes using AI to create a false representation of expertise - should be a big red flag as to the character of person using AI for self-aggrandisement.

What does this have to do with investing money for captives?

For investment advisers, reputation and trust are everything. New and inexperienced financial advisers  rely on the use of AI to answer RFPs and even build portfolios. But experience matters. I’ve written about this before, and I’ll keep saying it - AI won’t replace seasoned financial advisers.

Our strategist John Stoltzfus has been quoted many times saying: “Know what you own and why you own it.” I quote him frequently as that is also part of my mantra for asset management. A financial adviser who uses AI to build portfolios but doesn’t actually understand what it’s created risks ending up with a portfolio as attractive as Dr. Frankenstein’s monster – created by simply stitching together corpses then zapping them with electricity (read that as “AI”).

Long before the Large Language Models hit the scene, computer analysis has tried to outwit, outsmart, or simply outguess how to make money in the stock and bond market. Numerous apps and websites promise to build investment portfolios for (typically novice) investors based on age, income, monthly spending, and “risk tolerance” - a term so vague it’s almost meaningless.

Some advisers use similar systems to create investment models and portfolios for captive insurance companies. I know of an investment advisory company that has given up completely on active equity management stating “you can’t beat the index”, clearly missing the point that beating the index is rarely the stated goal of a captive insurance company. In fact, I find the IPS of almost all captives specifically states their goal as rates of return and liquidity that will allow them to obtain a reasonable rate of return with low volatility and the ability to meet all anticipated insurance payment obligations. Nowhere does it state “beat the S&P 500” or any specific index, for that matter.

Indexing vs benchmarking

Let’s take a brief detour to understand the difference between indexing and benchmarking. Indexing means owning an index - such as the S&P 500. But even then, you don’t actually beat the S&P 500 as the returns of the benchmark presume dividend reinvestment and exclude internal and external management fees. Additionally, the biggest downside to indexing is market swings. In the past 5 years, we have seen the S&P 500 (and many other indices) fall 20%, 30% or more in a short period of time. Most insurance companies can’t afford mark-to-market losses of that magnitude in their portfolios.

Benchmarking is a little more sophisticated. Customised benchmarks are calculated and employed to see how the asset allocation is performing relative to a blended benchmark that matches how the captive is actually invested.

In active asset management, experienced financial advisers employ separately managed accounts (SMAs) run by experienced managers with long track records of success in very specific areas of the stock and bond markets. One manager might focus on mid-cap value stocks, another on large-cap international, large-cap growth and so on.

Fixed income managers are equally focused and divvy up the bond market by government, municipal or corporate bonds and further parse it with allocations to US, European, Global, or emerging market debt all further sorted by credit ratings. No matter how you look at it, there are more funds and asset managers than there are individual stocks or bonds in which they can invest.

AI (Artificial Intelligence) investment models vs AI (actual intelligence) investment models

The tyro financial adviser will plug in parameters for a captive investment portfolio such as the investment amount, time horizon, risk tolerance and whatever asset allocations are allowed in the Investment Policy Statement. The computer will then kick out what funds to own and off they go.  This has historically been called “Garbage In, Garbage Out”:

“…if you put into the machine wrong figures, will the right answers come out?"... I am not able rightly to apprehend the kind of confusion of ideas that could provoke such a question.  Charles Babbage, Passages from the Life of a Philosopher

These models do not factor in thoughtful, experience-based prognostications of where stock valuations currently lie and where future valuations are likely heading. There is no substitute for experienced portfolio manager expertise on the fixed-income side with respect to the current level of interest rates as well as the anticipated level and direction of interest rates.

Absent - in these “tyro” models - are the “actual intelligence” used by experienced advisers and portfolio managers. They miss the nuances: upcoming known insurance settlements, analyses of frequency and severity of claims, or any attempt at asset-liability matching (ALM). That’s where actual intelligence - experience and insight - makes all the difference.

Increase the quality and quantity of information - and add to it a healthy dose of actual intelligence - the captive can feel confident that the likelihood of them hitting the objectives of their investment policy statement have been dramatically tilted in their favour. AI is a tool, not a solution or a substitute for judgment.

This article was written by Jack Meskunas, a financial adviser with Oppenheimer & Co. Inc. who can be reached at (203)975-2084 or jack.meskunas@opco.com. This article is not and is under no circumstances to be construed as an offer to sell or buy any securities. The information set forth herein has been derived from sources believed to be reliable and does not purport to be a complete analysis of market segments discussed. Opinions expressed herein are subject to change without notice and do not necessarily reflect those of the firm. Additional information is available upon request. Neither Oppenheimer & Co. Inc., nor any of its employees or affiliates, provides legal or tax advice.

Did you get value from this story?  Sign up to our free daily newsletters and get stories like this sent straight to your inbox.