Generative artificial intelligence (AI) is already changing the way in which scheme members are finding, consuming and engaging with financial information, a Quietroom whitepaper has argued.
The communication consultancy's whitepaper – How AI is changing the way your customers make decisions – said nearly everyone in the UK currently owns and uses at least one smartphone, tablet, or computer.
It said this means that one or more of three companies – Apple, Google, or Microsoft – were central to their lives, owning almost the entire UK market share of laptops, phones, tablets, search engines, web browsers and email apps.
Importantly, Quietroom said these firms were each "heavily invested in AI" and bringing AI to every single one of these experiences through Apple Intelligence, Google Gemini and Microsoft Copilot.
And it said there was currently "too little discussion – or even awareness – about the ways in which AI has already become part of people's lives".
Quietroom said it had tested real life financial services content with the devices, apps, search engines and AI assistants that customers are using and had identified three major shifts in how people interact with information through AI in the way they find, consume and engage with information – shifts that would have a significant impact on communications.
AI-powered search engines
The firm said the first key way in which AI was already affecting how people find information was with AI-powered search engines.
It said that, for example, in a traditional Google search, you enter your query, and Google presents a list of links, with the search engine expecting you to click through to these links to find the answer you need.
It said that AI-powered search changes this – noting Google was already starting to automatically generate AI answers to search queries. So, instead of sending a user to a website, Google reads the website instead, picks out the parts that are relevant to a user's question, then summarises, condenses and rewrites the content in the search window.
Quietroom said this means a user sees an answer that attributes you as the source, but which does not contain the exact content that you wrote and signed off – with users often taking the answers at face value even if the answer that Google gave them is wrong.
Indeed, Quietroom said these answers "can go wrong in lots of ways" – from omitting important details from a postal address to using sketchy sources or simply misunderstanding the content and getting it wrong.
It said testing how your content performs on AI-powered search engines like Google and Bing is now "essential" – noting it is only by testing that you can know AI interprets your content, who it associates you with, and whether you even appear in an answer in the first place.
The whitepaper added: "AI-generated answers are not deterministic, so you can never control exactly what they say. But if you can understand what they say, who they refer to, and where they go wrong, you can use this to start to refine your content to give AI the best chance possible to serve up accurate, useful and usable content for users."
How people consume your emails
Quietroom said AI was also reshaping the way in which people consume emails – noting that, for instance, Apple Intelligence had reshaped the way the users of Apple devices receive your emails through the new Apple Mail app.
It said Apple Intelligence now decides which emails are high priority. It looks at the email to decide how important it is, then it chooses whether to highlight it to a user as something they should look at.
Quietroom said Apple Intelligence also summarises the content of emails. When a user receives a notification saying they have received a message from you, they will not see the exact subject line and preheader that you wrote – instead, Apple Intelligence will read the message and summarise it in ten words.
The whitepaper said this meant users of Apple Intelligence no longer experience emails as they were written – controlling the first impression that your message leaves, and what points they find out about.
Quietroom said, as a result, messages that are carefully crafted can land very badly if you have not carefully tested and considered how they might be interpreted by AI when they arrive on a user's device.
It said: "It's easy to test your content. Simply find a user with Apple Intelligence and send them the email. Then, see what they get back… Not every user will see the same thing. But if you can at least test and plan for what some of your users will see, then you'll be far better prepared."
Interacting with information
The whitepaper said AI assistants are becoming increasingly commonplace – with Apple, Google, and Microsoft each allowing users to ask AI questions about any content they are viewing. With this, users can engage directly with documents or messages, receiving summaries or explanations based on their queries.
Quietroom said it had tested how AI summarised a number of defined benefit pension scheme retirement packs, considering how someone who is approaching retirement might use AI to interact with these packs.
Instead of reading the document, a user might ask ChatGPT, "which retirement option should I pick?" ChatGPT could summarise the options, suggesting the most popular choice and omitting subtleties in the original material.
When AI reinterprets your content, customers may act based on an altered version of your message. This could influence the decisions they make through AI-driven Q&As.
Reviewing content
The whitepaper concluded: "These changes to user behaviour are happening right now – and they're only going to become more prominent as AI adoption increases."
It said that, in this new world, users are less likely to visit your websites and content as you have written it; read your emails as you have written them; or read complex content.
To reduce the risks posed by AI interpretations, Quietroom said organisations should look to review and update their content to make it easier to read and use and remove outdated information and contradictions; ensure they write clearly and structure content effectively; and make sure they prioritise their users' needs, understanding what questions your audience has and create content to align with these needs.
It also said organisations should regularly review how AI tools interpret content, especially around calls to action and complex topics.
The whitepaper was based on a webinar presented by Quietroom's Thomas Joy and Russ Hope.
