The latest iteration of technological advances that either threaten or promise to take/do our jobs for us is artificial intelligence. Naturally I am interested in how that relates to me personally.
One incredibly important perspective would be to consider it for all the jobs I ask others to do for me. I am after all a consumer of billions of people and organizations’ labor but a supplier of just an insignificant few. That is a topic for another post.
In this post I will do what most of us do when faced with the prospect of progress in work—I will think about my own job—an almost always incomplete if not totally illogical framework. Try to remember this any time your defensive instincts kick in or the media writes a story concerning the evolution of work. Dynamism is the source of progress. The more secure a society is in the current status of its labor, the poorer that society is. Work is a cost. We want LESS of it.
“But what about jobs?!? We have to have jobs!”, I hear every politician and worry-minded voter say. Sure just keep running for the shelter of keeping everything the same. If that helps you get through the busy day, then great. But know that it is but a self-centered and regressive point of view that is constantly contradicted by everything else we all want and pursue—job competition for thee but not for me.
Speaking of me, I am a investment professional in a bank trust department. How is that for vague? Let me be more specific. I help manage the financial assets of a very wide range of clients: retirees, high-net-worth individuals and families, foundations and endowments, traditional trust accounts including beneficiaries, et al.
I manage marketable securities for these clients investing them as appropriate to help them reach their financial goals. I have over 20 years experience in finance, and I am a CFA® charterholder.
There are four basic ways I deliver value to clients. This is what they are paying me for:
Executing trades for basic portfolio management (raising funds, investing funds, rebalancing, etc.)
Communication with clients (hand holding, answering questions, giving performance reviews, etc.)
Developing investment strategy (figuring out how to invest and where to invest in general; e.g., underweight large cap growth, overweight small cap value, extend duration, et al.)
Financial planning (developing an investment plan specifically tailored to the client's needs and desires; i.e., setting appropriate goals within parameters of risk, determining constraints and adjusting recommendations accordingly, etc.)
I gave these in this specific order because this is the order in which I think I am replaceable by AI from easiest to hardest. Let’s look at each in more depth.
Executing Trades
This is almost a possibility today. It is for ~90% of trades. Simple rules and modeling can algorithmically accomplish this provided the rules are extensive enough. This would mean rules for nuances like tax implications (e.g., crossing trades within an asset class, avoiding wash sales, asset location given tax implications, etc.) and client preferences (e.g., asset modeling down to the client level—not all clients on a growth investment model of 75/25 actually want, say, a third of that 75% in international stocks).
The big leap AI could bring would be an ability to create, amend, and obey the rules without someone manually writing out all the rules the first time. The interpret/obey part is the only thing standing between a fully-automated modeled-account world for ~60% of clients today. This would be where a relationship manager or the client directly can ask the system (call it AI or call it simply a program) to “raise $10,000 for a distribution” or “invest excess cash” without an investment person acting as a middleman.
Today at almost all firms a portfolio manager is in that middleman position making sure it is done correctly. For the time being that is an important control feature that is value-adding to the client both in that it gets done literally correctly (e.g., $10,000 cash is liquidated from securities and made available for distribution) and it gets done optimally correctly (e.g., intelligently selling assets rather than just blindly adhering to the protocol). Eventually that middleman role will be a costly redundancy. That eventually is probably coming very soon.
Communication With Clients
This is a crowded category. It covers everything from basic information sharing to the very human role of talking clients off the (hopefully proverbial) ledge.
Obviously online access to view accounts has helped address the most basic information needs clients have. An AI advancement I can think of specific to this would be the ability for a client to ask an AI chat within an online access website direct questions like “how much income did my account earn last year?” and “where can I find a list of buy/sell transactions for February?” getting not just an answer but a helpful illustration of how that is found within the website and what it really means.
. . . Give a client the answer, he will hang up the phone for today. Teach a client to read a statement . . .
Beyond these basics we still need humans, right? Well, we already have examples of empathetic AI chats that have developed relationships with people including providing psychological help. They are at the least someone to talk to. The promise of developments in this space are ambitious, but the progress is equally impressive so far.
I expect how this would work best is where a chat program actually has a discussion with a client within an environment that is neither uncanny valley nor “I’m sorry, I didn’t get that. Please say a name or phrase for what I can help you with today . . .” 1
Responses can’t be canned answers even though most of the time the answers are obviously the same—and the client knows it. I can’t tell you how many times a worried client has called and more or less told me they knew that’s what I would say but they just needed to hear me say it. Beyond the reassuring words, though, is a person (representing a team) who actually is listening and who can back up his words with sound theory and historical proof. A successful conversation in these situations is some combination of what I am saying and how I am saying it.
Does it have to be in the flesh? Well, in some cases it does or it helps. But for a very large proportion of the cases it does not. I would always prefer a standard annual portfolio review be in person. There is something about sitting across from a person in the same room that is very hard to replace. Or rather something is lost when absent. But that loss is as often in my ability to read a situation and respond as it is in their ability to learn or get comfort from my presence.
The same can be said for the one-off request for understanding including cries for help. But most of these simply cannot be delayed until an in-person meeting can be scheduled. For these a phone call will have to do; therefore, the biggest apparent gap between AI and real-person intelligence is negated.
In this coming world give people credit to both understand the artificial status of the artificial person answering their concerns as well as relax enough to connect with it accepting the help it offers.
I would guess that 50% of my client communication is close to being replaceable in the next few years (technologically feasible but likely not cheap enough to be widely adopted). For the remainder most of that is coming but further out, and what doesn’t get replaced eventually is not for lack of ability but rather for lack of acceptance.
Developing Investment Strategy
There are two components to this. One is the global view a firm has toward investment. This is its philosophy and process. The other is its strategic and tactical approach and weightings. The former is essentially who the firm is while the latter is essentially how the firm does it.
Forming a firm’s investment philosophy might be helped by AI, but that help is naturally limited. Additionally, so too is the extent to which this is an on-going process. An investment firm’s philosophy really shouldn’t change, or if it does, that should be a slow, deliberate evolution where the speed and opaque nature of AI are detrimental if not dangerous.
Changing a firm’s investment positioning is where AI can play a more useful role. Still, in this area I see AI as a very helpful tool rather than a replacement. At the risk of making the same mistake the chess world made a few years ago predicting the hybrid human-computer player would always be superior to either by themselves, I will make a similar call. I think AI will reduce work and enhance competitive advantages (average is over), but it will not completely replace humans (me).
Financial Planning
This is the area of client-engagement where plans try to survive first contact with the enemy. This part is very close to the prior category of client communication. It is where we sit down with a client to develop a formal investment policy statement or a retirement plan to take two broad examples. This is all about listening, interpreting, guiding, and advising the client. It is where expectations are set for both parties.
The reason I placed it in its own category rather than the prior one is that I think this is actually the better fit. Although the same dynamics are at play, in this case the problem-solution is of another realm. Think of it as the difference between solving the autonomous driving problem and solving the where-should-I-drive problem. I (still) think that eventually we will have so-called driverless cars ubiquitously. Yet I don’t think they will determine for us where we should go.
To take it a step back, the difference could be more like route planning versus destination planning. Programs like Google Maps can rarely be beaten when determining how to get to, say, a specific restaurant. That example is analogous to the first two categories (trade execution and client communication). Here the problem is to determine where to eat and when to eat. Google can indeed make suggestions, but these are most effective when we already know a lot about the answer. In the case of financial planning, the ability of AI to help is limited to the extent clients know very specifically what they want to accomplish, what constraints they need to consider, what risks they want more/less exposure to, etc.
AI is therefore helpful when little help is needed and very dangerously unhelpful when help is most dear. I reserve the right to revise this, but I feel like this is a category where AI has much more limited replacement capability.
Conclusion
I very intentionally am not making strong predictions. The order of where I feel replacement opportunity lies is the point I want to make. This is true both in a more to less ordering as well as a sooner to later ordering.
I am fairly sanguine about this I hope because I am philosophically consistent—I embrace competition and creative destruction as positive forces, methods for good. However, I cannot escape that my attitude, perhaps reflected in the vague predictions above, is an optimistic spin on what indeed could be an acute threat to my short-term future. Still, I for one welcome the future where AI takes from me that which it can do better and offers to me an ability to do better in substantial other ways.
Bring it on.
very well put, especially the inclusion of the welcoming new overlords reference. :)