The pensions sector has historically lagged behind in adopting technological advances^[1]. Pensions administration plays a pivotal role in ensuring the financial wellbeing of scheme members and, therefore, requires efficient operations.

Artificial Intelligence (“AI”) can be referred to as machine learning, which mimics human intelligence while increasing efficiency. The adoption of AI in pensions administration is a topic of growing interest, and its implementation has sparked both enthusiasm and scepticism. There are undoubtedly processes in the pensions administration sector which AI can contribute to, but could the potential benefit outweigh the outlay?

Potential Benefits of Deploying AI in Pensions Administration 

Accuracy of Calculations:  The impeding Pensions Dashboard initiative necessitates the automation of pensions calculations.

There is a key opportunity for AI systems to process large volumes of data quickly and accurately. It will come as no surprise that AI can reduce the need for human operators, reduce errors, and increase efficiencies. Moreover, AI could also provide administrators and Trustees with summaries, KPI analysis, and data trends that can be used in a multitude of circumstances and tailored to a variety of stakeholders with minimal input.

Data: With many DB schemes progressing toward buyout, data rectification exercises are being undertaken in preparation for approaching the insurer market. AI could support administrators by reviewing data sets quickly, identifying inconsistencies, errors, and, in some cases rectifying the issues at the source.

Fraud Prevention: Pension scams pose a growing threat to pension scheme members and Trustees, particularly in the wake of the Covid-19 pandemic in which pension scams increased by 400 percent^[2].

AI tools could be implemented to successfully verify the identities of pension scheme members using methods such as biometrics, verifying identification documents and other advanced authentication processes. This could significantly reduce the risk of scammers impersonating a member to gain access to their pension pots.

Security: AI has the ability to analyse transactions in a manner which can recognise and identify any suspicious patterns or transactions, such as duplicate payments, unexpected transfers, suspicious behaviours and anomalies. This analysis could provide an opportunity to expose errors and issues that have previously gone unnoticed.

Additionally, AI can operate continuously, offering real time safeguarding whereas historically, outside of periodic audits, costly large-scale reviews would be necessary.

AI tools can also be implemented in other forms, such as monitoring behaviours, pattern analysis, language processing and anomaly detection with the added advantage of operating in real time, continuously.

The potential time and cost savings through real time monitoring is staggering alone. When considered alongside the further benefit of AI being unbiased, the economic advantages are even more apparent. 

Enhancing Member Experience: For many members, their pension will be the second largest investment following a property, therefore they should feel informed and empowered to manage their pension.

Many industries have already implemented ‘chatbots’ that operate 24 hours a day to respond to general queries. AI has the ability to learn the more it receives questions and interacts with both members and administrators. This means that the AI could be ‘trained’ to understand Scheme specific queries and responses will be personalised and consistent based on the underlying data.

As individuals have a multitude of apps at their fingertips, including finance, business and organisation apps amongst many more, there is an expectation for real time information and functionality to be available.

DC Pension administrators could seek to utilise AI in their platforms to review member records and provide them with tailored investment options based on their member’s wants. Members may not always be well informed on investment methods and could utilise AI’s research results to tailor their investment options.

The ability of AI in providing investment advice was showcased when a UK personal finance comparison website sought recommendations for a theoretical fund, comprising of 30 stock selections. Notably, the performance of this AI-led ‘fund’, eight weeks following the advice surpassed that of ten of the UK’s most favoured funds, with an impressive increase of 4.9 percent^[3].

Efficiencies and Fees: AI has the potential to complete routine tasks, respond to general queries, complete data entry, calculations, and document filing.

As previously alluded to, these automations could result in a reduction of human labour which, in turn, could lead to cost savings for pension providers. The pensions industry is currently suffering from a shortage of labour resources and the utilisation of AI could relieve pressures from administrators by completing routine tasks and handling general administration queries and calculations, allowing resources to be used in the most efficient manner.

Although infantile in the industry, the potential for AI being used to streamline processes and accuracies being improved is ever increasing.

No Human Bias: Due to the computational approach of AI, it approaches queries and analysis in an unbiased manner and follows the information it is provided with.

As a result, there is great potential to remove human bias from a number of processes resulting in reduced errors and further consistencies. These consistencies should only contribute to efficiency and cost savings within the industry.

Attracting Talent: Due to many job roles in the industry being innately repetitive, it is understandable that talent may seek more lucrative opportunities. However, the implementation of AI would allow administrators to focus on contributing value to a scheme. In addition, the use of AI in the industry would attract a number of individuals with varying skillsets, opening the industry to further potential modernisation.

Potential Pitfalls of deploying AI in Pensions Administration

Loss of Skilled Employees: With the industry suffering from a shortage of skilled workers, implementing AI may lead to administrators retraining or leaving the industry due to concerns that their jobs may be at risk. Suppliers and Trustees would need to consider the cost savings and potential efficiencies versus the potential skill loss^[4].

Data Security and Regulation: Given the sensitivity of pensions data, pensions administrators will want to ensure that AI complies with stringent data security and privacy policies. To date, the UK government has not released any regulations regarding AI but have created an AI task force^[5]. Without regulations, AI may be viewed as a risk to administrators and the handling of member data.  

Potential Bias: As AI operates on algorithms, it can carry biases through assumptions in which it has been trained and therefore requires oversight, careful planning, and human monitoring to prevent any discriminatory outcomes. Monitoring the output will also ensure that data being used or referenced is not skewed.

Risk Management: Trustees when reviewing implementing new processes or suppliers take into consideration they will improve or increase the risks to the Scheme. For many Trustees they may be unable to identify the risks that implementing AI may bring. Trustees may also want to wait to see what legislation and regulations are implemented regarding data protection and AI specifically.

Human Interaction: In certain circumstances, AI may not be the most appropriate approach, such as dealing with sensitive member cases, such as death. Members on occasion will require humans to provide compassion, empathy and a personal approach that currently cannot be replicated by AI. Administrators on occasion will deal with complex and unique cases that require human interpretation of laws and regulations to ensure compliance and a fair outcome is achieved.

Artificial Intelligence is becoming an integral part of our daily lives, such as Amazon product recommendations and smart speakers. However, perceptions about AI vary widely, and the scepticism extends to the pensions administration sector.

While there are clear benefits to integrating AI into pensions administration, there is a need for thorough evaluation, clear policies, and robust risk management to ensure its successful and responsible implementation in this crucial sector. Pensions administrators and Trustees should prioritise the best interests and experience of scheme members while embracing the potential modernisation and improvements that AI can bring.