What are the barriers, benefits and risks of rolling out artificial intelligence and related technology in pensions in the UK?

What are the barriers to rolling out artificial intelligence and related technology in pensions in the UK?

The biggest barrier for the introduction and roll out of artificial intelligence and wider pension technology solutions is the availability of high quality, accurate, standardised and digitised data. This is a challenge for many schemes and administrators, especially those that rely on legacy data.

Pensions are not at the cutting edge of big data and artificial intelligence. Compare the sophistication of online retailers with what is available in the pensions industry. Even within the narrower confines of financial services, pensions are behind banking and insurance in technological development. Open banking has revolutionised the way that banks provide services to their customers and we’ve seen the rise of online challenger banks and a range of financial technology (FinTech) service providers.

Society increasingly expects information to be available at their fingertips. Pensions can seem somewhat slow and behind the curve compared to this. Take the example of someone who has started their first job after university and has just been automatically enrolled into a workplace pension scheme. This individual will be used to accessing their bank accounts on their mobiles and are carrying increasingly complicated transactions using online and mobile apps. Pension savings can seem anachronistic in comparison (e.g. if you have processes that are carried out by post or online portals that are out of date and difficult to navigate).

What are the benefits of rolling out artificial intelligence and related technology for pension savers?

Pension dashboards will be the main push for standardised, high quality, accurate and digital data in the pensions industry. Once pension dashboards are a legal requirement, they will:

• provide the push for schemes and administrators to sort their data out;
• be portals for greater engagement, especially with younger savers; and
• produce innovation that flows on the back of the data that has been prepared and pushed out as part of pension dashboards.

For members, there could be:

• a much better online or mobile experience (including the convenience of having all pension savings in a single place via pensions dashboards);
• faster and more detailed information (including visual representations and modelling of information);
• more options, including the ability to carry out processes such as fund reviews and selection via mobile apps and online services;
• quicker resolution of queries and faster processing of requests; and
• efficiencies leading to cheaper administration costs.

Tied in with open banking, the rise of pension dashboards and the introduction of artificial intelligence into pension schemes will be the foundation for developing holistic approaches to financial wellbeing. Examples include:

• penny scraping (i.e. automatically rounding up each item of expenditure to the nearest pound and saving the difference in pensions or other savings);
• savings overflows (i.e. when people have excess money in their bank accounts at the end of the month and the system prompts them to save this or is set up to save this automatically); and
• prompts to increase pension savings when wages increase or there are one-off payments such as bonuses.

As well as the opportunity for engagement, education, financial advice and increased pension savings, there are costs and risks for the pensions industry. Achieving high quality, standardised and digitised data will be costly for some schemes. It may even be the final nudge that drives scheme consolidation.

What are the risks of rolling out artificial intelligence and related technology for pension savers?

Data that does not meet quality or industry requirements

Poor data could lead to poor outcomes for pension savers. Automated advice, nudges and modellers rely on high quality data to produce good outcomes. Many pension schemes still rely on legacy systems, even paper, card and microfiche records. Schemes and administrators will have to “invest heavily” in systems and processes to bring them up to the requirements for pensions dashboards. David Fairs, Executive Director for Regulatory Policy, Analysis and Advice at The Pensions Regulator has said that: “if the data behind the digital decision is inaccurate, the advice could be catastrophic for savers.”

Data protection and data and cyber security

Pension schemes hold a lot of personal and sensitive personal data. They are therefore attractive to hackers and vulnerable to attack. Ironically, those with paper records are better protected from attack. There are also broader issues connected to data protection and the sharing of data around systems as envisaged for pensions dashboards. There is also a danger that technology and systems are implemented ­quickly and ahead of regulation with regulators struggling to keep up.

Decision makers understanding systems and processes 

Trustees will remain responsible for the consequences of implementing new systems and processes. If they do not understand how these systems and processes work, they could find themselves exposed. Examples include algorithms that produce unintended consequences (e.g. a roboadviser that provides misleading advice to a particular group of pension savers because of an error in the underlying algorithm).

Understanding and being responsible for algorithms

Pension trustees, administrators and advisers should not expect solutions by putting in place algorithms that they do not understand. Regulatory bodies will still, ultimately, expect the decision makers to understand the systems and processes that they put in place. Decision makers will not be able to hide behind the argument that they didn’t understand the algorithm when it has an unintended consequence. 

An example of this comes from outside the pensions sphere. At the end of 2017, Air Berlin collapsed. This increased the demand for air travel amongst Air Berlin’s competitors. Lufthansa, one of Air Berlin’s main rivals, had a ticketing system that automatically responded to this increase in demand by increasing prices.

The German Federal Cartel Office “rubbished Lufthansa’s claim that recent price hikes on domestic routes were purely down to automated booking software, saying companies should not “hide behind algorithms””. The president of the Federal Cartel Office, Andreas Mundt, commented on the flag carrier’s argument, stating: “That’s beside the point. These algorithms aren’t written by dear God in heaven. Companies can’t hide behind algorithms.” In the pensions world, this principle of responsibility will apply to trustees and third party service providers.

Consolidation of advisers and administrators

The burden of achieving high quality, accurate, standardised and digital data will require heavy investment in systems and processes. This will drive consolidation but, when this applies to an already consolidated industry, it could result in a limitation of choice with a potential impact on the service levels and value for money that trustees can achieve.