PensionsJan 25 2022

Could a computer replace an actuary?

twitter-iconfacebook-iconlinkedin-iconmail-iconprint-icon
Search supported by
Could a computer replace an actuary?

The use of paper and post is also commonplace in financial services, and both factors prompt regular questions as to why the broad financial services industry seems to be such a laggard when it comes to adoption of basic technology, never mind the potential offered by machine learning and artificial intelligence.  

Machine learning is about the use of algorithms to interpret and manage large volumes of data, while AI is about automating decision making. 

In the pensions arena, decades of legacy issues are at the root of the problem, according to Sir Steve Webb, former pensions minister and now a partner at LCP.

Webb says one of the unique challenges associated with using capabilities such as machine learning in the pensions world is the quality of the data available.

Data dilemmas

He says: “Something like machine learning can be useful, but the data has to be good quality. And in the pensions world it often is not. For example, if you have someone who is presently 68, and begun working at 18, that is 50 years worth of data, which will have started off with the record kept on paper, and transferred potentially between employers, and then onto a computer, where there was the capacity for human error in the data entry, but also, over those years, there have been lots of updates to software systems etc, and errors can result from that.

"Then you have things like names, I mean, I’m Steve Webb, but on some systems I could be Steven with a V, or another Steve Webb could be, [or] Stephen with a PH, so at the moment the quality of the data is an issue, but that is fixable in time, and in future it will likely be different.”

As more of the workers who begin to draw down their pension began to work in a computerised age, the likelihood is the quality of the data will improve and enable machine learning to be used more frequently. 

In terms of the capacity for AI to impact how pensions work, one of the more obvious areas is in terms of the work of an actuary, much of which revolves around mathematics, where computing power should be able to lap that of the human.

Webb is a partner in a company of consulting actuaries, so perhaps it is understandable he is sceptical of the ability of AI to disrupt his industry, but he says there are also currently regulatory factors at play.

He says tasks such as calculating life spans and other such roles, where clients are segmented, could be automated, but for “the element that is individual to everyone, for that you need individual advice, which I don’t think can be replicated”.

Webb adds that while it is possible for a client or adviser to get a pension value today by computer, “for it to be a legally binding figure, and for the process to start, an actuary has to be involved”.

The role of regulation is also relevant to the deployment of machine learning in the pensions space, according to Sam Turner, senior consultant at Altus. 

He says that businesses are not legally obliged to accept digital signatures or letters as proof, and many still require paper copies. He adds that another challenge faced by advisers in this area is that, even when the pensions data on a client is held by one company, for example a life company and is accurate, the issue that comes up is that different business units within the company have different bits of the data, and with business units often moving around, it makes the task difficult for the adviser in a way that technology cannot really fix. 

Turner does say machine learning can help with the “adviser stuck on the phone problem”, as he feels the technology can be used to “triage” calls, that is, ensure that human advisers speak only to those who need to do so, with the rest dealt with by machines. 

Steph Willcox, head of actuarial implementation at Dynamic Planner, says: "The role of a pensions actuary hasn’t fundamentally changed in many years, and with very little chance to challenge or change the way that things have always been done, many of us have left the consulting world.

"As with any statutory role, the role of a pension scheme actuary is very unlikely to ever be made redundant by AI and machine learning and so the supporting roles are also unlikely to be able to be disrupted much.” 

But she adds there are ways in which technology can have an impact, saying: "Outside of traditional pensions consulting there is far more scope to involve machine learning and AI, particularly in the world of retirement planning. Financial advice, investment return modelling and cash flow planning all lend themselves to more modern takes on computing.

"No one wants to be given advice based on black box projections where there can be no explanation for the results shown other than 'computer says no', but utilising the power of computing at an earlier stage, perhaps with regards to stochastic modelling of asset returns, is a very viable and under-utilised area for exploration.

"It’s vitally important that results from any AI or machine learning can be challenged and interrogated, hence the need for caution in this area." 

Rise of the robots 

Webb says the limits of technology on the pensions landscape can be illustrated by examining the evolution of robo-advice companies, where the investment proposition is centralised around risk profiles, rather than personalised.

He says: “What you will find is, the new entrants to the market are able to do more because they don’t have the problem of legacy data and legacy systems.” 

Robert Cochran, retirement expert at Scottish Widows, says: “AI has itself taken an extremely important role in supporting our customers.

"Within the past two years we introduced our virtual assistant (Val), which is able to support people with their queries and provide tailor-made solutions 24 hours a day. In 2021 alone, almost 40,000 customers used Val, giving us a mountain of data to continuously improve the AI experience based on real-world interactions with the system.

"High advocacy rates of 70 per cent show the value our customers find in Val.

“The real excitement for the future capability of AI will come with the continued development of open banking, alongside open finance and [a] pensions dashboard. The three together will force financial institutions to think more like fintechs to enhance their offering, and build on their investment in robotics, AI and machine learning.”

David Thorpe is special projects editor at FTAdviser