I believe that most of us in the investment management world still do not truly understand artificial intelligence. What we often refer to as AI (and read about in the investment management media on a daily basis) is generally nothing more than intelligent rule-processing. Of course, nobody gets excited about intelligent rule-processing, whereas when you mention AI there is suddenly an air of expectation in the asset management boardroom.
To me, intelligence is applying knowledge to a set of circumstances and making a good decision. It’s not seeing the same set of circumstances again and repeating what worked last time – what some would call ‘machine learning’, where it doesn't matter how many times you repeat a process, you will always get the same answer.
In contrast, artificial intelligence has to regard, approach and solve problems in new ways. That includes being able to predict what will happen in a different set of circumstances or look for patterns. With AI, after a time we may start to receive different answers from the same inputs because it generally involves broader data sets and far higher volumes, with all the associated probabilities. After 10,000 occurrences the answer may be X, but after a million occurrences the answer may be Y.
With the proviso that we are all agreed that AI is not a precise definition but rather a loose term, then yes, I’m all in favour of more ‘AI’ in this industry. In the true, academic sense of the term, however, the issues we often discuss in terms of AI are really nothing of the sort.
Notwithstanding this lack of clarity concerning definitions or terminology, over the last few years I have come across some well-known asset management firms who have devoted much energy and resources in the pursuit of creating a ‘future state’ where everything in the investment chain - from investment strategy, product development, distribution, client acquisition, client onboarding, client servicing and so on – will be connected with AI throughout. Whilst a very exciting development, such a move does raise a number of questions.
The first question is, how much does it cost to build or maintain this kind of AI-driven environment? Certainly, this is never going to be cheap – there are costs and now substantial restrictions on asset managers holding vast amounts of data that identifies individual clients (unless those clients have expressly given their permission). Perhaps it could be argued that this approach would be easier for a larger firm to execute than a very small one.
My second question is, can it be said that AI is always a response to market demand or are there occasions when the technologists are giving the industry something that it isn’t asking for, with the belief that it will ultimately change everything? True AI it may not be appropriate in every situation.
In the world of investment reporting, for example, I am of the opinion that whilst automation has delivered wide-ranging benefits for processes, for the people who run those processes and for the end client, AI is still in its infancy. Any vision of a fully automated, people-less investment reporting process seems a long way beyond the horizon.
My third question is, does such a future state always make sense? Whenever I am faced with a totally automated process, I generally end up getting frustrated because I can’t get exactly what I want; I have to settle for the output the service provider wants me to have rather than what I would choose. Furthermore, whilst a machine should be better at spotting patterns than an individual, an individual has the capacity to provide the insights or the ‘lightbulb moment’ that a machine will never have.
On the other hand, a high degree of automation can bring significant benefits. As investment managers scale up, their businesses become more complex and their clients more diverse, the need to speed up the production and delivery of investment reports becomes ever more critical. Technology has been a vital enabler for the reporting process and the automation of investment reporting is becoming widespread.
Creating reports manually is by its very nature time-consuming. Automation will help reduce reporting risk, improve timelines and increase throughput without undue increases in fixed costs or headcount. It will also maintain reporting quality without compromising standards.
In investment reporting (as in most other front, middle or back office processes), to get exactly what an asset manager desires it needs to engage with a human being and there needs to be a degree of negotiation. In a cookie-cutter world, yes, AI would always be the answer for reporting, but really, is that how an asset management firm should view all of its investors? Are they really all the same? I’m not sure that investors would benefit from such a high degree of homogenisation.
Investment operations are not always black or white; there are sometimes a thousand shades in between.