The pensions world is known for its love of acronyms – GMP, TPR, LDI, CDC – but now there’s a new acronym on the block. AI, or artificial intelligence, has seen huge leaps forward in recent months and large language tools, like the trailblazing ChatGPT, have taken the world by storm. As we begin to grasp the possibilities – and risks – of this innovative new technology, the pensions industry is beginning to consider how it could be applied in the context of defined benefit (DB) pension schemes.
As pension lawyers, we certainly have thoughts on AI’s implications for the legal landscape and the potential legal risks that could arise from its use (and misuse). But pension lawyers are also curious by nature, so we decided to try an experiment: we asked ChatGPT to summarise the potential implications of AI for DB pension schemes.
Read on to find out what it said (in the box below), followed by our three key takeaways from its fascinating response.
Our three key takeaways
1. Bias can creep in
We didn’t direct ChatGPT to take a particular stance on the question, and perhaps unsurprisingly, its generated article was generally very supportive of the impact of AI. While some legal risks and ethical considerations are mentioned, the overall message is focused on potential efficiencies and benefits.
In pensions, neutrality is key. Conflicts of interest must be carefully managed, and communications with members must be carefully drafted so as not to be perceived as advising on or influencing member decisions. There is clear susceptibility to bias in AI tools, demonstrating the importance of a critical human eye in reviewing any output.
2. AI doesn’t think of everything
The generated article drew out several potential uses of AI technology in a pensions context, but there are other possibilities it has overlooked. One example that comes to mind is member complaints, whether through an internal dispute resolution procedure (IDRP) or to the Pensions Ombudsman (TPO).
While there is some support available to members, the reality is that members are often unfamiliar with the legal and technical framework of DB pensions and find themselves unable to express their complaint clearly. This contrasts with trustees, who are well-supported by a team of expert advisers with experience in managing and responding to such complaints. As the internet is flooded with tales of overturned parking fines and consumer victories, all thanks to a well-crafted AI letter, could ChatGPT be a tool for members bringing IDRP and TPO complaints?
ChatGPT also failed to mention a notable downside of AI: its environmental impact. Training large language models like ChatGPT uses significant energy and water resources, and trustees need to be mindful of this as part of their overall ESG strategy.
3. Sometimes pensions law is just too complicated
AI may be capable of making transformative leaps in science and innovation, but apparently it is no match for the notoriously complex topic of anti-franking of guaranteed minimum pension increases. We asked multiple times, in multiple ways, for a simple summary of the requirements, but ChatGPT remained completely stumped. Pension lawyers 1, AI nil.
So what does it all mean? It’s clear that we are only just beginning to understand what this rapidly-evolving technology is capable of, and no doubt the DB world will be well-placed to capitalise on its potential, while being mindful of its risks. We’re curious to see where AI goes from here, but when it comes to complex pensions issues, it seems there will be a place for humans for a while yet.