Financial services is an umbrella term for a vast industry with some of the biggest global companies in the world, which deal with some of the public’s most at-risk datasets and assets. 

These traditional institutions have spent years, decades, even centuries building consumer trust by ensuring the public’s money is secure, their financial investments are sound, and their assets are future-proofed. 

Financial services regularly deal with people on the worst days of their lives, often dealing with people’s greatest vulnerabilities in turbulent economic times, as trade undulates and the cost of living crisis squeezes everyday consumers. 

FS institutions do not treat this burden lightly – they are often known in the tech world for their cumbersome legacy systems and mass data sets which need to be wrangled across siloed systems. Their conservative, stable approach to innovation is mean to instill trust from their customers, but times are changing.

The rapid rise of AI and the growing push for AI investment, adoption, and deployment is putting pressure on a more conservative industry, and rightly so. 

The advent of AI offers a number of risks to a more-risk adverse industry dealing with the products of people’s livelihoods, who are tasked with approving and managing mortgage loans, dealing with emotional and monetary consequences of inheritance, advising financial planning for pensioners and first-time parents. 

But AI is here, and is already being adopted by financial institutions, pushed across sectors and systems with vigour and at a pace that seems to show no sign of slowing. 

However, considering the stakes of FS, speakers at DIGIT’s inaugural FS Technology Summit in Edinburgh urged the hundreds of delegates in attendance to show pause, caution, and critique when initiating AI into their organisations. 

“Things are moving extremely fast, so you have to be able to be agile, adapt, and innovate, and that’s particularly hard in financial services organisations,” Andy McMahon, principle AI and MLOps Engineer at Barlcays. 

“We have this need to manage risk and build trust,” he said. 

Jumping on the AI bandwagon is simply not a viable option for FS firms, as the risks and lack of trust this can incur is too high a price to pay. 

But the expectations for AI are incredibly high, research from analytics firms McKinsey and Bloomberg says. The GDP of the UK would be added to the global economy annually due to GenAI through the 2020s and into the 2030s, McKinsey predicted. 

Similarly, Bloomberg predicted that 10% to 12% of all technology spending in organisations would be funnelled towards generative AI. 

McMahon is critical, he admitted, especially of what this would look like in reality. 

“Any of you managing IT budgets for your organisation: Can you imagine carving 10% of that budget, everything you’re looking at – and to put it on a nascent new technology that’s just hit the scene? That’s quite dramatic!” 

McMahon is encouraging others to be “very critical” of investments in AI, of this almost all-encompassing drive for AI adoption, this fear of being left behind, of disengaging investors, of losing out to competitors. 

When implementing AI into a firm, McMahon said it needed to be treated like investing in any other new technology – it should not, and in fact, cannot be simply shoved into a system.

Back to basics 

Looking at the firms that have most successfully integrated AI into their systems looks a lot different than offering an AI chatbot for basic customer service queries – generative AI is at its core, a general purpose technology that can be applied to a range of domains. 

FInding where a firm can capitalise on this technology across a specific vertical “and not just following the hype” differentiates a firm from those jumping on the bandwagon. 

“Giving everyone an AI assistant isn’t going to automatically drive ROI,” McMahon said. “What’s going to drive ROI is when you drill down into your specific verticals within your organisation. We’ll have such specific questions for our organisation, and you need to find where the value of AI translates for you.” 

So, how does one identify a positive use case for AI? 

McMahon suggests keeping to the six key abilities of AI as a central framework for figuring out if it makes a compelling adoption case. 

These include: search, summarise, analyse, generate, translation, and transcription. 

“If it’s not one of these six with GenAI, then I’m just trying to put a square peg in a round hole,” he said. 

It’s also important to go beyond these basic principles – just because AI can summarise things, does not mean there’s a business case for this integration. 

AI is also touted as a productivity improver, but again, McMahon is critical of the data. 

“People are saying ‘I am ten times more productive than I was’ since adopting AI, but I want to challenge this – where’s the data?” 

Is this just a feeling, or are people really more productive?

“Maybe they’re faster; they don’t necessarily mean they’re really putting out better quality code now.”

It’s important to note that McMahon is not bashing AI – he’s simply being critical of half-hearted implementation efforts amongst companies, who throw AI into their systems without proper forethought or a sound enough foundational basis. 

And in reality, firms are still waiting at the starting line when it comes to the AI race. 

A report from BCG found that around 40% of companies across various sectors had yet to take any action on AI adoption, while only 10% were at the stage where they could scale. Half were at the planning phase in their AI journey. The rush may be more talk and promises than reality. 

More than that, it shows that there’s a strong barrier to translating enthusiasm for AI into actual return on investment. 

Ensuring that this is achievable truly does bring us back down to the basics of being AI ready, and just business ready – “good existing capabilities around your data around your foundations, your infrastructure, your processes, and you are basically a strong user of predictive AI analytics already,” McMahon described the perfect candidate for scaling AI. 

Organisations just at the start of their digital transformation journey will likely struggle to implement AI effectively without this basis. 

“You’re going to have to think very strategically about how you make the right investments,” McMahon said. 

Recommended reading

• Data Security Holding Back Growth for UK Financial Services Firms
• Futures Institute Launches Compassion in Financial Services Hub
• UK Finance Chiefs Bet Big on AI – Is The Workforce Ready?

Building Trust By Not Trusting AI

Financial services should know by now that trust takes time to build trust with customers. The same should be true for trusting new systems, especially when it comes to AI. 

“Don’t assume that the off the shelf solution is just going to solve what you need to do,” McMahon argued.

AI skills also play a role – ensuring that staff are capable of using AI to better their work, not just complete it, can make a world of difference. 

Research from Git Clear shows that AI-generated code from Copilots have lead to an increase in duplicated codes and eroding code quality. While workers may be getting more done, the quality of this work may be decreasing. 

Experts in their field may be able to use AI, spot any errors, and fix them, but this also points to the requirement of having a strong foundational understanding and ability to 

AI models have bias, they have the ability to hallucinate, and the ability to mislead. Understanding what you want from an AI model, and if what they produce is quality, is key. 

“It’s just that balance between the chase, the hype, “give everyone an AI co-pilot and assume I’ve got huge ROI”  or actually critically evaluate,” he said. 

“We have to go back to basics and just say what is the business case, what is the opportunity I’m chasing? What are the metrics that tell me I’m succeeding. What are they for?”

Determining a use case, creating metrics to measure this return – not just on what people feel, but what is actually happening in terms of productivity – is vital in ensuring an AI investment is done responsibly. 

Even when trust in AI is built responsibly, and the hype means an IT department gains the funding for their AI transformation or integration, trust still needs to be built. 

“We’re still gonna face the same challenge, which is when I implement this stuff, we’re going to have a bit. Of a reckoning where. You know the business. The organisation is quite rightly going to say where the value show me the money, right.”

While McMahon has found boards heartily encouraging him to invest in AI, organisations will still have questions about return on investment – and feeling more productive might not cut it. 

“The worst thing that we can do as organisations is spend millions, in some cases hundreds of millions of dollars, implementing technology. And then find out years later that it would have been better to wait.”

Having the right metrics and measures in place is essential for an organisation to trust that their AI implementation and investment is worth it down the line. 

This is, however, easier said than done. The AI skills gap is currently a top worry for firms as they try to implement the new technology, and measuring data and metrics of its return on investment can be tricky to disseminate and explain to a board of directors that also may not fully understand the technology. 

McMahon suggests some sort of central AI department – an AI centre of excellence, for instance – that can focus various departmental integration initiatives, explain and track the progress of AI projects, and reflect these back to leaders. 

This can help financial services organisations maintain a balance between innovation whilst maintaining trust within their organisation and with their customers.

“Legacy financial services firms have quite rightly built a reputation and trust for being conservative,” McMahon concedes. “But now, we’re in an era where we have to break from some of that. I’ve never seen anything move this fast, so we have to try and think: ‘What is the balance we are going to strike?’”