Artificial intelligence (AI) has quickly gone from being a far sighted fantasy to integral to the running of many industries, including financial services.
Yet, for this to be a sustainable development, there needs to be some resistance to the large amounts of hype around AI.
This is an excerpt from The Future of AI in Financial Services 2025 report, which was a special edition for the inaugural Finextra event,
NextGen AI. Click here to read the report.
Firstly we’re going to look at the best approaches for banks to investigate which use cases work best for them, before seeing which use cases were highlighted by contributors as some of the key use cases they are seeing.
AI use cases: slow and steady wins the race
With the popularity of AI growing, many financial institutions will be feeling pressure to dive head first. However, before going gung ho, banks may want to consider exploring which use cases will give them the best value and usage.
Bahadir Yilmaz, chief analytics officer, ING, emphasised the use of a “structured approach” for financial institutions investigating and identifying AI use cases. He further added that this approach should align with “risk and regulatory constraints and
prioritise ethics.”
Other banks we spoke with shared this view for a coordinated method to internal research. Graham Smith, head of data science and innovation, NatWest Group, said: “Taking use cases from pilot to product requires the right skills, processes and governance
to have the greatest impact.”
Smith described NatWest’s bank wide programmes investigating their best use of AI to help their “customers, stakeholders and businesses thrive”, and an exercise in 2023 established 100 potential use cases, which they narrowed down based on priority.
From ING’s side, Yilmaz stressed the safety and security of AI uses, he said:
“AI models based on personal data should be free from bias, explainable, transparent, responsible, and always based on consent. Data should not be used beyond its purpose and therefore, clear regulation regarding data retention is key. There is no concrete,
ubiquitous regulation yet on how AI models should work. However, financial institutions have a duty and responsibility to set their own model ethics framework.”
Yilmaz explained that ING have a 20-step process to evaluate every AI system for 140 risks, and only after that would they be allowed to go into production. He added: “Applying GenAI in one business problem is only 5% of the job, the other 95% starts after
that when you are building all the systems around it to make it safe, secure and non-biased.”
A challenge for some financial institutions is not to get wrapped up in the hype, Pavel Goldman-Kalaydin, head of AI and ML, Sumsub warned against jumping on the latest AI trends: “By starting with specific pain points and aligning AI with these core objectives,
they’ll avoid the trap of chasing tech for tech’s sake and instead create real value.”
Isa Goksu, CTO Globant UKI and DE, gave practical advice on how to approach investigating AI in financial services, starting with “comprehensive assessments to map current AI usage.” He continued: “Establishing and investing in governance and risk management
frameworks tailored to AI is key; if you don’t have clear guidelines on responsible use right from the start, you’ll struggle to implement effectively further down the line.”
However, contrary to this slower pace, Shaun Hurst, principal regulatory adviser at Smarsh, stated: “Financial firms could adopt a faster approach to building and testing new ideas.”
Hurst did emphasise balancing this with “strict security protocols” and added: “The best approach would be to bring different people together who understand banking and those who understand tech.”
Mirroring Hurst’s point, Goldman-Kalaydin also pointed to creating a “cross-functional” team: “AI projects don’t just need data scientists and engineers; they need people who understand the business deeply, like compliance officers, risk managers, and customer
service leaders. This combination helps ensure AI solutions are grounded in the institution’s real-world needs and operational realities.”
From ING’s experience, Yilmaz advised: “Financial Institutions should also launch pilot programmes and start small with prototypes to test AI solutions before full deployment.”
This was something also suggested by Goldman-Kalaydin, who said starting small with pilot programmes can save on making “huge, resource-heavy investments right from the start.”
Goldman-Kalaydin additionally emphasised the importance of data quality and governance, he added: “AI can’t deliver on its potential with poor-quality data. In finance, this isn’t just about getting better results but rather about avoiding costly errors
and regulatory pitfalls. Strong data governance practices, along with clear policies on data privacy and security are crucial.”
With all of this in mind, financial institutions should be looking at taking a structured approach to their AI investigations, which takes a census of their current usage and pain points. Here are some of the key use cases with interviewees proposed.
AI to fight financial crime
Lord Christopher Holmes, Baron Holmes of Richmond, highlighted fraud detection and prevention as a use case to focus on, he said: “There is no question, we are living in an epidemic of fraud right now. AI, in combination with other technologies, always human
led, offers the best opportunity to address this desperate challenge.”
Fraud has seen some increases, with the Office of National Statistics (ONS)
reporting a 19% increase in consumer and retail fraud incidents in 2024 from 2023 in the UK. Adding to this pressure for banks is an increased number real time payments and the large amounts of real time data.
Hurst observed that some financial institutions are already “leveraging AI to prevent fraud, pre-empt cyber-attacks and navigate the complex regulatory environment.”
He added that AI has the ability to analyse “vast amounts of real-time data to identify suspicious patterns, prevent fraudulent activities and detect emerging cyber threats.”
Given the increasing pressure to tackle financial crime, the preventative use cases of AI are something which Goldman-Kalaydin noted. He said: “Traditional fraud prevention methods often led to frustrating false alarms, flagging legitimate transactions as
suspicious. Now, with AI learning each user’s unique behaviours, the system becomes almost personal—tailoring itself to individual transaction habits. This personalisation means fewer false positives and a much smoother customer experience. It’s a win-win:
customers experience fewer interruptions, and banks spend less time on unnecessary investigations.”
Adding to this, Goldman-Kalaydin discussed natural language processing (NLP) and its abilities in behavioural analysis, he said: “This is particularly valuable as fraudsters become more sophisticated in their communication tactics.”
Goldman-Kalaydin also pointed to AI’s ability to uncover fraud networks: “This goes beyond catching an isolated fraudster; it’s about revealing the complex networks behind large-scale fraud. Graph analytics is a powerful tool here, mapping out relationships
and spotting hidden connections that could indicate collusion. It’s remarkable to think of AI systems piecing together these intricate webs, flagging fraud on a scale that would be nearly impossible for humans to uncover alone.”
Overall a benefit for fraud detection and prevention teams is increasing efficiency, as Hurst said: “It also helps legal, risk and compliance teams stay ahead of regulatory changes and review requirements.”
AI improving efficiency internally and externally
Fraud teams are not the only area where AI can improve efficiency, Goksu claimed we will be seeing more agentic AI workflows soon, a kind of AI which can independently make decisions. He explained: “AI is transforming the lending process by automating credit
scoring and loan approvals, increasing efficiency and reducing approval times.”
For Goksu, AI could provide greater front end efficiencies, especially in the area of credit checks: “In the next five-10 years, credit risk analysis will fundamentally change as AI-augmented services transform social engagement models across industries.
This shift will allow financial institutions to leverage a broader spectrum of data, including social behaviours and interactions, providing a much more comprehensive and accurate assessment of creditworthiness.”
Smith said that from their use of AI they are seeing an increase in their “efficiency, productivity, and overall colleague satisfaction.”
He reported that they are saving 15 minutes per call on in their private bank because “relationship managers have been using call summarisation tools with their clients to capture details, summarise the call and extract key facts, which gives the relationship
manager greater freedom to focus on the customer during the call.”
However, Smith also argued that the internal efficiencies offered by AI is something they are focusing on as they “become a simpler and more efficient bank.”
He elaborated: “We’ve equipped our HR colleagues with AI tools as they support colleagues across the bank with their everyday HR queries. ‘Ask Archie’, our 24/7 AI chatbot for HR queries, now uses Gen AI to simplify colleague experiences, creating better,
more natural conversations. This in turn enables colleagues to access information quicker, whilst our human agents can spend more time working on more complex support cases.”
Hurst also pointed to AI’s use for efficiency in internal sustainability teams, who are required to report on which investments are green and a breakdown of what type of green investment their loan falls into. He explained: “The amount of data required for
this is vast, but AI can process large amounts of data at a rapid pace, which will help financial institutions to classify their green investments and make more informed decisions around sustainability.”
Customer interactions and personalisation
Efficiency as a bank can be greatly impactful in improving customer interactions, as Hurst explained: “Improving customer experience remains a key AI use case in financial services, with conversational AI being introduced to simulate human interaction.”
Building on these improved customer interactions is the ability for personalisation. Smith stated: “AI will enable levels of personalised engagement like we’ve never seen before, empowering us to predict customer needs whilst delivering simpler and more
engaging interactions with customers.”
Goksu posited the potential of AI for personalisation in the next era: “Banking products in the next era will be completely driven by AI and they will be hyper personalised. It also enables precise, tailored financial planning, aligning with clients' unique
financial goals.”
Smith gave the example of their AI-powered chatbot Cora, which has generative AI integrated: “Since the launch of that pilot in June, we’ve seen a 150% increase in customer satisfaction, and a halving in the number of cases requiring colleagues to intervene.”
Robo-advisors and automated trading systems were another area Hurst saw as ripe for AI, giving both efficiency and personalisation, however, he warned: “These applications require robust oversight and risk management frameworks given their potential impact
on market stability and individual portfolios.”
As well as enabling personalisation, Smith spoke to the overall strength of the banking industry in leveraging AI, as it “has both huge amounts of data and the ongoing need to evolve its customer proposition, and so the opportunities for AI to enhance service
levels and efficiencies is perhaps more significant than a lot of other sectors.”
Moving forward with AI use cases
“Ultimately, exploring any new use case needs to align with a bank’s strategic goals, so for us we’re looking at how AI can simplify how we operate, become more efficient and effective, and make it easier for customers to deal with us,” commented Smith.
Many financial institutions are seeing common themes in AI use cases which could be beneficial across the industry. Banks should maintain a balanced and careful exploration of which work best for them and their pain points.