Emma Storr
Content Marketing Manager, MoneyNext
Generative AI is revolutionising hyper-personalisation in banking. However, challenges remain.
Richard Keirnan, Global Head of AI Platforms at NatWest Group, explores hyper-personalisation with genAI with leading financial services events provider, MoneyNext.
GenAI vs traditional methods
Automation, dynamic content generation and multimodal inputs set generative AI apart from traditional personalisation methods. Unlike static rule-based approaches, generative AI adapts in real time, enhancing engagement and efficiency. “There are a few things which are real game changers; one is a contextual understanding,” highlights Keirnan. “It’s not that we’re just understanding the words; we’re truly understanding the whole thing in context.”
With AI-driven automation, content creation is more scalable, surpassing traditional natural language processing (NLP) capabilities. “If you asked me a year ago: could we retrain the large language model for personalisation? I would have said it’s categorically too expensive. Now, on commodity hardware, it is possible,” enthuses Keirnan.
Banks must sanitise data and
implement strong bias mitigation tools.
Key challenges
Despite promising advancements, challenges persist with concerns centred around data security and ethical AI use. “If you’re a regulated organisation and you start using a large language model, it is possible that it will start learning from your data and allowing that data to leak is not ok, so you do need the right architecture,” emphasises Keirnan.
Banks must sanitise data and implement strong bias mitigation tools to prevent this from happening. “We need to make sure that we get this right and this goes back to the right monitoring tools,” offers Keirnan. “You need guardrail tools. You need to be sure that we are monitoring bias at model level, but we’re also testing it.”
Emerging trends
As the technology continues to advance, Keirnan predicts a shift toward more focused models to aid in enhancing accuracy and minimising errors. “Large language models may well turn into SLMs (small language models) — not small, but specialised,” states Keirnan. “We’ll merge the technologies together and get an output that’s actually far more advanced, more accurate with less parameter hallucination.” Over the next five years, AI-driven personalisation will be essential in banking and the wider financial services industry, making interactions more intuitive, secure and efficient.
Discover more unmissable insights from industry leaders and the latest event updates from MoneyNext (wearemoneynext.com)
