Problems AI can solve in banking

The banking industry has been quite resistant to change ever since its beginnings. But financial services are not immune to the biggest technological revolution the word has attested caused by AI. Apart from being under pressure to adapt to the digital economy, banks have started to discover some really valuable AI use cases. AI has made progressive inroads in the financial sector and is reshaping banks’ approach to their people, processes and data. From customer service automation with chatbots, security, fraud prevention and detection, all the way to internal process optimisation, AI is transforming the traditional way banks work. 

Mattias Fras working as a Group Head of AI Strategy & Acceleration at Nordea gave his perspective on the problems AI helps solve in the banking industry at Data Innovation Summit.

AI in Nordea

Nordea was a frontrunner with technology and as early as 2015 they started making first steps with robotics. Mattias had been a big part of Nordea’s breaking new ground in automation and robotics. From the first robot going live in 2017 until the time of  Mattias’s presentation at the Data Innovation Summit in March 2019, Nordea had 300 robots within their operations. The next stage was building chatbots which now support around 120,000 conversations with customers and employees per month. Following the latest tech trends, Nordea is now focusing on AI.

Mattias admits that although everyone is talking about AI, banks and financial institutions, in general, are not doing that much with AI, although the potential is huge. As Mattias works on a group level in Nordea, he has an overview of all departments and is convinced that all functions will benefit from the new technology. “For the banks, it’s going to be like the electricity and it’s going to change everything,” states Mattias.

As a result of Nordea’s launch into AI, they released their project Nordea AI, which according to Mattias is a pretty bold move for an old bank. Their AI project is Nordea’s testimony that they are serious with embracing AI and believe in the inevitable transformation of banks. “And we also want to attract the biggest problems to solve in the bank,” resolutely says Mattias.

For the banks, [AI] it’s going to be like the electricity and it’s going to change everything.

The main areas where Nordea has applied AI are:

• Pattern recognition (finding irregularities in data)
• Foresight (predicting)
• Customisation
• Decision-making 
• Interaction with customers

Nordea AI couldn’t happen sooner, Mattias claims, because of several occurrences that are affecting banks:

• Regulations – There are many new regulations introduced in the last couple of years referring to the use and handling of data and they’re putting pressure on banks to comply.
• Business models and ecosystems are changing
• Customer behaviour is changing
• New competition enters the market, such as tech and fintech companies 
• The technology is developing at an unprecedented pace.

We want to attract the biggest problems to solve in the bank.

What they learnt in the process of AI adoption

As one of the pioneers with AI in banking, Mattias points out several things they have learnt in the process.

1. AI is not about buying a CRM or an ERP system or another piece of software and hoping it will solve your problems. It’s about being open to exploring and ready to take the “red pill”, in Mattias’s words. Nobody knows how banking will look like in 5 or 10 years time. In order to find out, you must be ready to take the plunge.
2. The banking industry is all about regulation, risk mitigation and trust. And it should be like that because they are working with finances. But unless we are willing to swap our shiny shoes with snickers i.e. changing the way we think and reimagining what we can do with data and machine learning, we won’t see the change we want to see.
3. AI and machine learning transformation are about starting small and learning fast. Small use cases help prove the value and gain trust in the applications. And over time, the applications are put in production and lead to new game-changing solutions. But it’s important to step out of the rule-based policy-way of thinking and start trusting the data-driven approach, which is a huge step for a bank Mattias emphasises.

Picking the right problems to solve in banking

When looking for which problems to fix with AI, Mattias brings up the 10X projects, which refer to exponential projects with AI. And that should be the focus of every AI project according to him. As a comparison, 99% of AI projects within banks focus on improving the current processes, products and CX. But Mattias likes to go further, reimagine how a complete process would look like and build it from the ground up. Because “there is nothing more inefficient than improving upon something you shouldn’t be doing in the first place”, claims Mattias. So for that reason, his mission is to combine high ambition, service design and new technology in 10X when deciding what problems should be solved.

3 areas where Nordea uses AI

The three main areas where Nordea applies AI for improvement are:

• Intelligent digital customer experience – Nordea is providing better, more customised digital customer experience. They offer their customers with a virtual assistant named Nova introduced in 2017. With 120,000 interactions per month, it’s changing the way customers interact with the bank. 
• Giving employees superpowers – Nordea is reducing manual tasks for employees, enabling data-driven insights and automating decisions. Two examples which Mattias gives are:
  • 1) Using AI trade surveillance (TS) to discover patterns that are impossible for the human employee to discern among thousands of transactions. They use unsupervised clustering models to detect irregularities. The challenges they are facing are few real cases and limited historical data. But the benefits are of course valuable such as better detection rate and building learnings and trust for the next generation TS AI
  • 2) Using AI for anti-money laundering (AML). This use case combines human and AI efforts to improve efficiency by prioritising alerts before they are sent to customers. Fortunately, they have many use cases and enough historical data to use in the model. The benefits are more time to focus on important AML cases and refine training data for next-generation AML AI.

• Instant banking – AI focused on claim handling process. Nordea’s instant claims reinvent the time-consuming insurance process lasting 2-3 months with low transparency, unstructured data and human judgement. Their instant banking process has a rule engine using ML that detects if there is an insurance claim, hands it to a robot, which in turn sends it to a pattern recognition engine to make sure there is no financial fraud and pays out the money. The end-to-end process is cut to only one day. The process has a few additional steps for doctor’s notes: they are scanned with an OCR, made in a structured text, handed to a text analytics engine that looks for keywords and medical history, and ultimately makes the claim decision. For more complicated cases, they also have a human agent involved.

There is nothing more inefficient than improving upon something you shouldn’t be doing in the first place.

How banks can transform with AI

In the end, AI. machine learning and models serve only as enablers and components for solving problems. What should be the ultimate goal, Mattias suggest is banks to embrace a modular way of thinking and combine AI applications (microservices) to solve multiple problems across the entire value chain and different use cases.

To be able to do this, departments working with AI within banks need to step a bit outside of how they usually do business to avoid being stuck in processes. Mattias gives the association of a start-up within a bank that solves a particular problem.

Challenges to tackle in banking

Some current challenges that offer room for improvement with AI in banking are:

• Changing the mindset and the way problems are solved. Not improving on current processes, but reinventing them.
• Changing internal processes
• Changing the idea of what kind of people and talent are necessary. Finding people that can solve problems, and not sticking to internal processes and procedures.