[Skip to content]

Sign up for our daily newsletter
The Actuary The magazine of the Institute & Faculty of Actuaries
s
.

Adapt or adopt

05 FEBRUARY 2020 | ELLIOTT COX

Elliott Cox asks whether improvements to advanced learning will push actuaries out of work

web_p36_student-illo_jan_feb_2020

During the past decade, technology has transformed the way we work. Calculations that were previously completed by hand can now be run through Excel models to produce reliable answers in a fraction of the time, with the actuary’s role shifting to independently checking and verifying these results. I’m sure very few of us would complain about this; we can now spend our time on more interesting work, refining actuarial techniques and developing new solutions for clients. 


As the industry continues to develop, however, will previously accepted methods become too ‘primitive’ and imprecise to justify? Computers and modellers become more sophisticated as the general workforce gains experience in model-building, and it’s now commonplace for workers in the financial sector to use powerful software and tools that are capable of much more than we currently use them for. In other areas, we’re seeing the introduction of ‘robo-advisors’, which provide personalised advice by following mathematical rules or algorithms. The asset management market has led in this space, given the limited impact of ‘softer facts’ that fall into this advice, but it’s conceivable that future developments will widen the range of behaviour we can model and code.


Do we really need these ‘overpowered’ tools, though? In some cases, the processes we rely on today are as precise as we would reasonably need them to be – and where there is limited precision, it’s often a conscious choice. Much of what we do takes account of the available data, and where we don’t or can’t know exact values (for example, financial data at future dates), we make assumptions based on our current expectations. No amount of modelling will remove the inherent randomness of the future events that affect these figures, and unless we can use advanced learning tools to observe trends in the data that we haven’t identified from conventional modelling, there is little to gain from overcomplicating things. It might be one of the few examples where an actuary accepts the adage ‘if it ain’t broke, don’t fix it’.


How would advanced learning tools impact errors and omissions? We would expect reduced human error to be a benefit, but this relies on a robust underlying model, which is only possible if an appropriately experienced person can correctly create these tools. Will it be necessary for actuaries to have coding experience, or will actuaries be replaced by non-actuarial coders to meet this need? We’re beginning to see statistical modelling appear in the earlier exams, but there is still a long way to go before these tools will be commonplace in our work. Actuaries aren’t coders by nature, and there will always be different generations of actuaries that have qualified under different syllabuses. 


At its heart, our work relies on producing sensible numbers – and questioning why the results we have come to could be justifiable when things look strange. This latter aspect is  something an automated process cannot do. What do you do when the sense check breaks down?


Separately, many of us work as advisers to clients. Consultancy is deeply entrenched in human interaction, and involves building relationships, keeping up with topical (not necessarily ‘codeable’) developments, cutting through the details of technical information, and relaying all of this in a digestible manner to ensure clients receive the solutions they need. When we over-rely on modelling, there is arguably no recourse for recipients of our work when the process breaks down. It seems reasonable that there would be market apprehension  around the reliability of such new tools – will clients trust significant advice from a robo-advisor if they can’t ‘see the whites of their eyes’?


Ultimately, though, we would expect technological improvements to translate to efficiencies. Better models, reduced human involvement and tools that can easily reproduce results (without the need for detailed expertise) will undoubtedly cut costs. Pressures from the market dictate that we need to continually adapt and improve if we are to provide a better service to those we work for and with.


Will future technological innovations revolutionise the industry? In the longer term, quite possibly. There would be a lot of detail to work through in the short term, not least convincing users of our work that the technology is trustworthy – would you be comfortable receiving pensions advice from a robo-advisor, with technology in its current state? Furthermore, will actuaries be replaced by coding experts or online tools? It’s perhaps more likely that the focus of our work will shift, with some actuaries primarily spending their time building new processes and others remaining in more ‘traditional’ advisory or specialist roles.


Elliott Cox is a student editor.


Configure your Portal