Landscape in the press: January 2024
Our client services director Johnathon Ryder recently spoke to Pensions Age about the challenges that member communications face with the emergence of AI. The piece — titled ‘Member communications: Is AI ready to step up for pensions?’ — has just been published in the January edition of their magazine. Here’s Johnathon’s piece in full:
Simplifying complexity is crucial for understanding and engaging with pensions, as well as delivering targeted, tailored, and timely messages to drive action. Traditionally, this has been done by humans, but is it time for a change?
AI is omnipresent
King Charles III addressed the topic in his November speech, emphasising the need for a gradual adoption with adequate safeguards. Elon Musk believes it could spell the end of humanity. It has even been named the ‘word of the year’ by Collins Dictionary. There is no escaping it. AI is causing divisions in various industries. The 2023 writer’s strike in Hollywood, lasting 148 days, was the longest in its history. Writer’s fear being replaced, and AI is even taking on the role of artists by generating artwork, images, music, songs, and data analysis.
What can we currently do with AI that is easily accessible and usable?
Quite a lot. Web tools like OpenAI’s ChatGPT and Dall-E, Microsoft’s Copilot, and phone apps like NovaAI offer a range of capabilities. According to ChatGPT, AI can assist with creativity and art, education, and automation. In essence, it can design a simple logo, create a newsletter, write an article, support learning, and perform data entry tasks. However, there are certain issues and considerations to keep in mind.
What is Landscape’s experience with using generative AI?
Our experience with AI has been a mixed bag, but we continue to explore its potential whilst understanding the risks. We challenged ChatGPT to write a page explaining Pension Increase Exchange (PIE). Unfortunately, the output we received was fundamentally flawed and contained incorrect information. However, a second experiment to draft an article explaining the basics of inflation, how it is measured, and why it is important was more successful. With adjustments to match our writing style, it was accurate, correct, and saved time.
Generalists and specialists
This led us to consider the need for both generalists (graduates and junior team members) and specialists. Our four-eye review process involves a generalist conducting research, understanding the subject matter, and creating a first draft. The specialist then reviews it for technical accuracy, compliance with pensions rules and regulations, and ensures the accuracy of any figures. In the AI experiments, AI replaced the generalist in creating first drafts, which the specialist needed to review and edit. The inflation article required minimal changes, while the PIE article needed almost a complete rewrite.
Providing feedback to AI is currently challenging
Specialists often spend more time correcting the first draft instead of having the generalist do it. This means no real savings are achieved. The impact on generalists and specialists is another consideration. With the traditional process, generalists gain knowledge and understanding through the process, eventually becoming specialists themselves. With the AI approach, this knowledge transfer is lost.
So, are things ready to change?
Well, yes and no. For schemes with smaller budgets, AI tools can potentially serve as starting points for articles, reports, newsletters, etc. However, thorough checks for accuracy and compliance are necessary. AI can generate images to accompany the articles and even create a podcast version using generative voice AI tools. Landscape believes that the process should be more collaborative between the generalist and AI. If the generalist can use AI to create a first draft and then conduct the initial check, the output will be more accurate when it reaches the specialist. This approach can potentially reduce the amount of work needed while knowledge is still being acquired and future specialists are being trained. That’s where we currently stand, but things are changing every day. In fact, this article is probably already outdated!
