The rapid advancement in artificial intelligence over the last two years is already impacting the workplace pensions market, with providers and consultants anticipating more significant transformation in the years ahead.
TO DOWNLOAD A PDF COPY OF THE ROUNDTABLE SUPPLEMENT CLICK HERE
This was one of the main topics of conversation at a recent roundtable event, hosted by Corporate Adviser, looking at the opportunities lying ahead for the industry this decade, and the challenges this disruptive technology will present.
Royal London’s group head of data strategy Lee Wilson made it clear that discussions around AI’s role in the pension industry are about developments happening now, rather than blue-sky future-gazing.
“AI is already here. We’ve been using more narrow forms of AI, like machine learning and first-generation language models, for some time in the financial services sector to add value for businesses. We are already starting to adopt generative AI tools to find ways to improve products and services in the pensions space.”
Much of AI’s use to date has been to improve and streamline internal processes, rather than on member-facing technology.
But Wilson said that the next generation of AI tools will transform the industry. “When I think about the scale of disruption in this space, it’s akin to computers coming into the world of business, or the shift that occurred with the advent of the internet.”
Wilson said one of the key changes happening now was the “democratisation” of AI tools.
This is opening up AI to a wider audience who don’t necessarily have coding or tech backgrounds, which is helping drive innovation. She added that newer AI tools are increasingly sophisticated, making them appear more intelligent.
“The latest generation of AI models can deliberately do ‘slow thinking’. In other words the model can explain what it is doing, or thinking at each stage, so users can interrogate that and audit it and understand what is happening in there.”
AI solutions are also increasingly multimodal, she says. “This is just a fancy way of saying they can deal with lots of different types of data. It doesn’t all have to be structured data in rows or columns; it can be free text, voice data or images, and it can generate content in all these forms too.”
Pension applications
Wilson said that these various aspects will mean AI can deliver better information for consumers and digitalised guidance — particularly as part of the proposed new targeted support services. “This will be a key opportunity for the industry,” she said.
“More ‘intelligent’ AI models — often called agents — are already far more effective than the chatbots of old. We’re getting to the point where a dialogue with the agent will feel like a conversation with a human. It will be very, very difficult to tell the difference unless the AI shares this information.”
The consultants attending the roundtable event said this raised some obvious ethical issues.
Howden Employee Benefits and Wellbeing head of pensions Emma Hadley said: “It is a good thing if these new models can help people have more productive and effective conversations around pensions and retirement. But to me, one key aspect is making sure people know that they are talking to an AI agent. I’ve seen all sorts of avatars where it looks like you are talking to a real person. What responsibility is there on providers to disclose this information?”
Gallagher Benefit Services senior DC consultant Rupert Redman added: “There
is a big danger that individuals will be misled by their own use of AI. It’s an exciting opportunity at the moment, we need to embrace these changes, but it’s a Wild West out there at present and I think the regulators need to get on top of this.”
Wilson agreed this should be made clear, and said that it was imperative firms had good governance around new AI models. Royal London had employed an ‘AI ethicist’ to address some of these issues, she said. However, while larger established providers may seek to adopt best practice on these issues, there were concerns that this would not be true of all AI providers, and many advisers want to see clearer regulation around this issue.
Accuracy concerns
For the consultants in the room, the other big concern at present remains the accuracy of content generated by AI models.
Hadley pointed out that many search engines, like Google, already embed AI into searches. “For more general questions the answers given are around 88 per cent accurate at present. But if you start asking more detailed or technical questions the accuracy rates are quite poor at present.”
She added: “We have had members who’ve used AI tools and got completely the wrong answer. This is an issue the industry has to grapple with, because this technology isn’t going away.”
Wilson said that providers like Royal London are building their own proprietary AI systems, trained on their own source material IP. “We’ve engineered them so they are Royal London-specific. We’ve given them knowledge artefacts which include the relevant regulations for our industry, information about our products and services, and even the company’s tone of voice and house style.”
This more rigorous approach, she said, can deliver far more accurate and specific answers to more technical questions. “The challenge for the industry is to create something that has the convenience and appeal of ChatGPT or a Google search, but the veracity is much, much higher.
“I think the knowledge management space is going to be a really crucial area and it will be important to develop skills and governance around that.”
While most consultants were sceptical about whether AI-driven solutions would replace individual financial advice, many saw huge potential in guidance and targeted support solutions, delivering a more effective service than previous ‘robo-advice’ options.
Omny strategic partnerships consultant Roy McLoughlin said: “As advisers there aren’t enough of us to go around, and we don’t see many younger people entering the market, so this is where this technology may help, particularly to those who are not accessing advice or guidance services at present.”
As well as having a more intuitive and ‘human-like’ interface — those at the event were optimistic that these AI-led agents could deliver more personalised answers to those seeking help with pension saving and retirement options.
Data challenge
But Wilson pointed out that in order to deliver this, pension companies have to make better use of data. Wilson has a background in retail banking, and has seen first-hand the transformation that initiatives like open banking had on the industry.
She said pension providers do not have the same rich data points as the retail banks, but could make better use of the data they have on members, while also incentivising them to provide relevant details about their circumstances.
“We need data in order to offer something more personalised and useful back. Customers aren’t going to share data if they think this is only going to benefit the provider. We need to demonstrate how this benefits the customer, by delivering better solutions for them, tailored to their needs. Employees — and the employers — need to see the value proposition and know that the more they engage with these tools, the more personalised the content will be.”
AI can improve internal processes to help facilitate this, she added, for example taking information from phone conversations. “There are definitely opportunities to take the data out of that interaction and deploy it in order to improve customer service through all stages of the journey. It could highlight vulnerabilities for example, and may ultimately remove the need for a lot of form-filling.”
Consultants said that this shows the potential for AI across the industry when it comes to delivering more efficient systems, not just for providers but for the consultancy side of the business too. Foster Denovo head of client consulting Gavin Zaprzala-Banks said: “Most organisations are starting to use AI internally to do many of those frustrating, time-consuming tasks. AI is helping make humans more human, freeing up time to enable us to do the genuine consulting and deep thinking that the industry needs, rather than just cranking the handle and keeping things moving.”
He said consultancy firms already make widespread use of AI through programmes like Microsoft Copilot, which is having a “real difference” on what teams can deliver and at what speeds.
Delivery dates
McLoughlin said the big question is when we will see more transformational change — particularly in relation to AI agents delivering highly personalised solutions — and whether this will mean better pension outcomes for savers? Did the AI experts in the room expect this within a couple of years, or by the end of the decade? Wilson said this was very much a near-term project. “The technology exists now. It’s the engineering and how people use it — that’s where the challenge lies.”
The current regulatory framework is potentially also an issue when it comes to rolling this tech out at scale at present, with Wilson noting one of the bigger challenges, when delivering personalised guidance, was “ensuring it doesn’t slip into advice”. However, Royal London said that it was in favour of the more “principles-based approach” from the regulators, which was allowing innovation at present.
Many in the room thought that AI would certainly deliver real benefits — but most thought there would still be a critical role for consultants and advisers, not least in highlighting the need to engage with pension issues in the first place.
Royal London’s director of policy Jamie Jenkins said in many ways this will be an evolution rather than a revolution. AI will be driving improvements in products and services, and people will be using these tools, but may not necessarily be aware of it.
He drew a parallel with changes in the banking industry in the last decade, where a combination of technology and better data sharing has facilitated faster account transfers and speedier payments. “People didn’t say they were using open banking, they have simply got used to making instant payments through the apps on their phone,” he added.
Jenkins said AI presents a huge opportunity for the pension industry to do things better, particularly when considered in relation to wider regulatory and technological change — be it the advent of targeted support, the consolidation of small pots, the introduction of retirement defaults, and most importantly, the launch of the pensions dashboard.