In this interview, we speak with Dr. Robert Murphy, Chief Economist at infineo, a company working at the intersection of AI, blockchain, and financial services. With a background in Austrian economics, Robert brings a distinct perspective to how emerging technologies can reshape the $3 trillion life insurance industry and beyond. He discusses the philosophical and technical hurdles of digitizing legacy financial systems, how AI can enhance rather than replace human advisors, and why blockchain may redefine trust and transparency. The conversation also explores the mindset shifts business leaders need as AI and blockchain transform how we think about value, risk, and ownership.
Explore more interviews here: Adam Famularo, CEO at WorkFusion — Leadership, AI Digital Workers, GenAI Challenges, AI Evolution, Risk Mitigation, Scaling AI, and Human Oversight
Dr. Murphy, how has your foundation in Austrian economics specifically shaped the way you design AI-driven financial products at infineo?
To be clear, our entire team contributes to the design of infineo’s financial products, but it’s certainly true that my input largely derives from my background as an Austrian School economist. The work of Friedrich Hayek (who won the Nobel Prize in economics in 1974 for his work on business cycles) in particular anticipated developments in both cryptocurrency and AI. Specifically, after spending the 1930s and 1940s devoted to work in theoretical economics, Hayek broadened his horizons with the publication of The Sensory Order in 1952.
This book adopted the Hayekian view of the market economy, which relies on undirected, local feedback mechanisms that exhibit a deep “order” at a macro scale, one that no individual in the system planned, and applied it to the human nervous system, yielding a sophisticated “theory of mind.” Scholars familiar with Hayek’s pioneering work have argued that it anticipated later developments in computer science and artificial intelligence, specifically. It is not surprising to an Austrian economist that large language models (LLMs) can exhibit seemingly intelligent behavior even though no component of the system “understands” the data, because, as Hayek demonstrated, there is a sense in which human beings exhibit rational action even though no individual cell in their bodies “thinks.”
Digitizing a $ 3 trillion-plus legacy life insurance industry is no small feat. What were the biggest philosophical or technical challenges you faced when blending traditional insurance principles with AI and blockchain?
The biggest technical problem we’ve faced so far is the heterogeneity of life insurance products in the tradfi marketplace. Just the seemingly trivial task of training our AI engines to “read” uploaded PDFs of user policies was surprisingly challenging, because there are so many product types (Universal Life vs. Whole Life vs. term) with various riders.
Additionally, each carrier may have idiosyncratic terminology for the various features of a given product type, and their presentation of projected policy performance may be formatted uniquely. Consequently, we had to obtain samples from every carrier we wanted our system to recognize, and we had many meetings to discuss the “theory” of how our system, which had to collapse the various carrier frameworks into one box in our ecosystem, should classify various items. The whole process was far more complicated than just, say, developing a taxonomy for Treasury bonds.
From your perspective as an economist, how can AI enhance, not replace, the human element in financial advisory and insurance services?
Here again, I can draw on my background as an Austrian School economist, where one of the major historical controversies was Ludwig von Mises’ critique of socialism. In the 1920s, Mises argued that without market prices, socialist central planners wouldn’t know the relative economic value of different types of resources (acres of farmland, barrels of crude oil, tons of iron ore, an hour of a dentist’s time, etc.).
Only with genuine market prices, Mises argued, could entrepreneurs assess, even after the fact, if they had transformed scarce inputs into goods and services that were more valuable to society. With the advent of computers, defenders of socialism argued that Mises’ critique was obsolete, because now central planners could rely on rapid machine computations to “solve the equations” characterizing an economy.
However, I have long pushed back against this claim by pointing out that the computers themselves are resources embedded in the broad market economy. By continuing to rely on market prices and the profit-and-loss steering mechanisms, it “frees up” the supercomputers to help entrepreneurs find oil deposits or design a more efficient airplane wing. Likewise with AI, it’s wrong to view them in competition with humans. The invention of a tractor didn’t put farmers out of work; instead, it freed up some humans to move out of agriculture into something else, because now a given worker, who remained a farmer, could produce fantastically more output per hour.
There will be bumps in the road, of course, but AI in the financial sector (and insurance specifically) will empower the best advisors to take on more clients and to provide more rigorous financial projections. Furthermore, when it comes to private financial data, some people might feel more comfortable “discussing” their situation with an impersonal AI engine such as the “AI Bob” that we have at the infineo website rather than someone who works at their company.
Could you share a pivotal instance where an Austrian School economic theory directly influenced a key AI strategy or product design decision at infineo?
As we continue with our internal modeling and product design for a blockchain-based ecosystem for life insurance products and derivatives, the rest of the team and I frequently invoke the principle that we aren’t central planners. If we are going to tokenize a particular element tied to a life insurance policy (or pool of such policies), we can, of course, come up with a theoretical value for it, based on “the fundamentals.” But ultimately, we are designing our ecosystem to rely on market prices, knowing that ultimately, a token is worth whatever the market believes it is worth. Perhaps ironically, this humility will make our system more robust during volatile times when many people might conclude that the market price is “wrong.” Say what you will about them, but crashes can happen, and we are building with that in mind.
Blockchain is often hailed for transparency. In the sensitive realm of life insurance, how does infineo leverage blockchain to build client trust in ways that traditional systems couldn’t?
When it comes to building financial products that involve “pooling” of components of individual life insurance policies, there is an inherent tradeoff: On the one hand, outside buyers will obviously want as much information about the characteristics of the constituent policies. (For example, someone might not want to construct a portfolio of policies all issued by the same carrier.)
On the other hand, users of our ecosystem who want to enhance the performance of their original policy through our infrastructure do not want sensitive medical facts put on a public blockchain. Modern blockchain technology (with robust AI-driven systems overseeing the procedures) allows us to strike the right balance in a way that would have been impractical in a traditional setting.
When introducing AI-driven financial tools to clients steeped in traditional processes, what was their greatest initial skepticism, and how did you address it?
Probably the biggest source of skepticism thus far has been the gap between AI hype and reality. Some people have the idea that a computer can’t make mistakes or at least can’t make mistakes when it comes to “objective” issues like financial products, and then, when they catch an LLM saying something wrong, they doubt everything else it generates.
We have tried to minimize these negative experiences by first, troubleshooting on our end and anticipating possible problems, and second, by managing expectations and showing users the proper way to interact with (say) our AI Bob engine to get the best results. For example, rather than simply asking, “What is the best type of insurance policy?”, which is extremely open-ended and in a sense an impossible question to answer instead the user could say, “Let me explain my current financial position and my goals, and then please give me the pros and cons for using various types of insurance products to achieve them.”
You’ve written for lay audiences about complex economic concepts. How has this experience informed your approach to designing user-centric, intuitive digital financial products at infineo?
As you say, much of what I’ve done in my role as a “public economist” is take complex topics (such as monetary or banking theory) and explain them in plain English, using examples that resonate with the average person. (“Why should I care about this?”) This aspect of my career has translated into what we do at infineo, where the underlying financial theory and calculations have to be corrected, but we also need to present the results intuitively for our end users.
After all, there are plenty of “facts” our dashboard could show someone about his or her life insurance policy, so it’s our job to design the AI interface and graphical displays to focus on what is likely most relevant for the individual user.
What mindset shifts do you believe are essential for business leaders, especially in legacy industries, when integrating AI into their core financial strategies?
I think the most fundamental problem people are making when it comes to AI is that they expect to type in a simple command such as “Give me a new product to launch next quarter”, and get a showstopping answer. Part of the reason for this is that certain AI engines really did pull this off; for example, some of the more recent photo and video generation is truly amazing.
But when it comes to incorporating AI into a legacy business, I think the leadership team should first familiarize themselves with the general capabilities of the latest versions of the various engines. For example, until you’ve had several “conversations” with an LLM, where you’re not just kicking the tires but you’re really trying to get help with a problem or task, you don’t understand just how nuanced and capable of error-correction they are. So I would recommend that executives first spend some time understanding the sense in which the latest LLMs can, in a sense, mimic remote workers who are quite reliable and work for a very low hourly wage—and then figure out how to pull out certain tasks from the company workflow to delegate.
Can you describe a moment when applying AI to financial data at infineo revealed an economic ‘blind spot’ that traditional models had overlooked?
I was recently using “Claude” to help me run Monte Carlo simulations, because I was writing up an explainer piece showing why pooling life insurance assets made for more predictable returns than holding the assets in isolation. I had a particular intuition about how the result would look, and I was guiding Claude to run the massive simulations just to get some actual statistics to give specificity to my hunches.
My main ideas turned out to be validated (phew!), but when Claude was commenting on the results, “he” pointed out some patterns that I hadn’t planned on seeing. So this was clearly an example where Claude let me do something that, five years ago, I probably would have just kept as my intuition and written up in words. Yet now, I could very quickly have the computer write up a computer program to bang out the simulations to illustrate my verbal claims, and then even give me commentary on the results and give a “second opinion.”
Looking ahead, how do you envision AI and blockchain reshaping not just the infrastructure of finance but fundamentally altering how people perceive ownership, value, and risk?
In financial economics, as it developed from the 1950s, there was a growing recognition that the value and risk of an asset couldn’t be defined in isolation, but only in relation to the investor’s overall portfolio. (This is why people care about a particular stock’s covariation with the overall S&P500, for example, and talk about its “beta.”)
In this context, I see the growth of blockchain-based finance, “governed” by AI engines that can perform decentralized analyses and execute transactions, will integrate the global assets and liabilities in a way that reduces inefficiencies and provides more predictable outcomes for every participant. For example, changing forecasts of longevity in various populations scattered across the planet will have implications for various types of assets beyond life insurance, and thus exposure to blocks of life insurance can be spliced and/or recombined in different packages to hedge and allocate the risk to the parties who are most willing to bear it. Done properly with the help of blockchain and AI, the result will be more affordable life insurance for everyone, and more predictability for people running nursing facilities and pension plans.
