Open banking in the age of AI.

With ChatGPT sparking a surge of interest in AI, we take a look at how machine learning models are changing the future of banking.

A wave of artificial intelligence stories hit the headlines this year after OpenAI made its ChatGPT language model available to the public at the end of last year.

The latest iteration of this impressive chatbot can pass college exams with flying colors and is already being put to work in the financial services sector. At Klarna, for example, it is providing a highly personalized and intuitive shopping experience by suggesting curated product recommendations to users who ask the platform for shopping advice and inspiration, along with links to shop those products via Klarna’s search and compare tool. And Morgan Stanley’s wealth management unit, is using ChatGPT-4 to help its staff find information quickly. Instead of scouring internal repositories for the latest analyst insights on a specific topic, staff can now simply ask GPT-4 and get an immediate response.

Search engines from Google and Microsoft (which have invested billions of dollars in OpenAI) are rolling out AI-powered search results that will allow broad public access to this powerful technology.

Indeed, the pace of development has even led to calls for a pause on advanced AI research. As with any new technology, there are risks associated with AI, many of which have been highlighted recently. In our tests at a recent hackathon, we found that ChatGPT produced inconsistent and unreliable results. For these reasons, financial regulators are closely monitoring the use of AI models, while recognizing their potential value.

Advanced language models such as ChatGPT represent progress towards general AI, but narrower machine-learning techniques are already changing the way we interact with technology on a daily basis, including in open banking.

Let’s take a look at how AI is improving the open banking customer experience, enhancing security, and increasing efficiency — and how it may be used in the future.

Open banking is perfectly placed to benefit from the rapid advances in AI because it’s built around APIs (application programming interfaces), which allow access to a large amount of financial data. This is essential for training AI algorithms to get better recommendations and predictions.

Designed to make banking more competitive, more accessible, and more secure, the EU’s open banking rules force banks to give customers more control over who they share their banking information with. This has given third parties access to a treasure trove of data that is ripe for AI to use.

By training a computer on this type of large dataset, AI techniques can be used to build a model and learn to identify patterns and relationships in the data, which can then be used to make predictions or decisions. Ultimately, this will result in better and cheaper services for customers.

Before we go further, here’s a definition of AI provided by ChatGPT:

“AI is the field of computer science and engineering that deals with the creation of intelligent machines that can perform tasks that typically require human intelligence, such as visual perception, speech recognition, decision-making, and language translation.”

AI is set to become even more impressive as computing power continues to increase — from focused challenges such as image recognition to tasks that require much more complex decision-making, such as self-driving vehicles.

Onboarding

From the first interaction with open banking, AI is already working away in the background to deliver a smooth experience. AI-powered identity verification allows users to open accounts in just a few minutes, completely transforming the onboarding process. 

• Document verification – Using the camera function on a smartphone, AI models can scan identity documents and verify them using the in-built security features.

• Live video biometrics – AI models can also perform facial recognition, comparing live video with ID document photos.

These are transformative use cases. They save time and hassle for customers, massively improving the user experience, while also delivering results that are cheaper and more accurate for the businesses that are using them. This ultimately leads to much better conversion rates and lower customer acquisition costs.

AI-powered identity verification can also be scaled easily, significantly reducing the obstacles that can get in the way of expansion for growth-focused businesses.

Categorization

One of the most valued features of open banking is the ability to categorize transactions. This seemingly simple step allows customers to see their finances in a new way, with spending broken down into useful buckets, such as grocery shopping, takeout food, clothes, and entertainment, for example.

In practice, this is anything but simple. Bank APIs don’t provide this extra information, which means that third-party providers need to add it afterward, and this is where AI can be particularly useful. Machine learning algorithms can be trained on large amounts of data to recognize patterns and establish relationships between different types of transactions and categories.

For example, an AI algorithm can be trained on a dataset of transactions that have already been categorized into different categories. Using AI in this way is still in development, but there is significant potential for these types of models to be used in open banking in the future.

Fraud detection

While effective identity verification improves the customer experience, it also helps to cut fraud and make open banking safer. Beyond this, AI also promises to help spot fraudulent transactions in real time. By learning to recognize patterns in transactions, AI can add an extra layer of security and give human analysts more time to investigate suspicious activity.

Regulators around the world are already recognizing the power of AI in fighting financial crime:

• In 2022, Singapore’s MAS published a series of white papers to guide the responsible use of AI by financial institutions, setting out key principles: fairness, ethics, accountability, and transparency.

• Also in 2022, the FCA in the UK published a report on best practices for the safe adoption of AI in financial services.

• As long ago as 2018, US federal regulators encouraged financial institutions to consider whether AI may help them to “better manage money laundering and terrorist financing risks while reducing the cost of compliance”.

Similarly, regulators in the Netherlands, Germany, and France have also weighed in on the subject, signaling a growing acceptance of new technologies in the fight against financial crime.

Advances in this area will see the development of cost-effective AI solutions that promise to make open banking even more secure. Financial criminals won’t stop developing new ways to circumvent regulations, including the use of their own AI-powered tools, so it’s important that banks and other financial institutions continue to upgrade their defenses.

While there will continue to be a role for rules-based protections and human oversight, AI has the power to spot fraud quickly and reliably.

There are many other ways that AI will be deployed in the open banking sector, including those that aren’t yet evident. Large language models such as ChatGPT will soon be helping to power a new generation of customer service chatbots, while AI models will get even better at predicting customer behavior and applying this knowledge to customer retention.

Without a doubt, digital-first players in the open banking space will be embracing these types of solutions to deliver next-level services — just as Kosma already is.

Ultimately, success in the AI age will be determined by how well businesses harness these technologies.