AI: Driving transformation in financial services

AI-driven solutions in financial services

AI-driven solutions are already commonplace in the world’s biggest banks; in the MENA region in recent years, we’ve seen Abu Dhabi Commercial Bank in Egypt using AI to improve its customer service and fraud detection and National Bank of Kuwait introduce an AI-powered chatbot. Fintech companies around the world – including Zest AI, a California-based developer of underwriting software which uses AI to assess the creditworthiness of borrowers, and the UK’s Bud Financial, an open banking platform that uses AI to help users budget and track their spending – are driving innovation and partnering with some of the world’s biggest institutions.

As AI technology develops, innovation accelerates. Large language models (LLMs) are the latest exciting iteration of machine learning and a form of AI that is particularly adept at mimicking human intelligence. It is widely agreed that LLMs have the potential to transform the financial services sector through their ability to analyze huge data sets, identify patterns, and make accurate predictions.

LLMs, such as the popular GPT-3 launched by OpenAI in 2020, are designed to understand, generate, and manipulate human language with an unprecedented level of complexity. Each generation of the model is being trained on larger and larger data sets, allowing them to understand more contexts and make them more relevant to more areas of knowledge.

LLMs developed in the GCC region are already competing against models built by tech giants such as Google. Falcon 40B, the UAE’s first large-scale, open-source AI model (developed by the Technology Innovation Institute (TII) in Abu Dhabi), was ranked the top LLM model by Hugging Face, a US company considered to be the definitive independent verifier of AI models, in the first week of its launch in May of this year. Falcon 40B can be used to create text and solve problems, and can be applied in customer service chatbots, translation, and sentiment analysis. TII believes that its focus on data quality means that Falcon outperforms GPT-3 for only 75% of the training compute budget. Another recent exciting development is Jais, the world’s most advanced open-source Arabic LLM – the bilingual model (developed by the UAE AI company G42, the US AI company Cerebras, and the Mohamed bin Zayed University of AI) was launched in August 2023.

Maximizing the value of your data

We see real potential in LLMs, particularly in their ability to mine the maximum value from an organization’s data. This is what’s behind our recently-launched offering Nortal Tark (‘Tark’ being the Estonian word for ‘smart’), an AI solution that uses LLMs to help our clients extract value from their data on their own terms. Or in other words, it combines AI’s capacity to acquire knowledge with the ability to do something useful with that knowledge.

In practice, this means that organizations can generate immediate insights from large amounts of their data. One of Nortal Tark’s core elements is Intelligent Search, which we install securely within the client’s controlled environment, giving them full control of data and privacy. The client then has the ability to interrogate all of the data that it has gathered, including emails, reports, memos and documents. The model will understand a question in any language and summarize the answer in a matter of seconds.

Significant potential with some challenges

LLMs have significant potential, and not just for the largest multinational financial institutions – they are a cost-effective and accessible solution. Meta AI developed LLaMa, for example, doesn’t need cloud access or exceptional hardware, and open models are available that are already trained to provide a high level of functionality. LLM chatbots such as Alpaca-LoRA and GPT4ALL are powerful, fast and efficient and can be used for customer service, content creation, and education.

There are, however, some challenges that organizations need to be alert to:

Security

While AI is helping financial services firms in their fight against fraud, that same technology is being used in increasingly inventive ways by hackers and fraudsters. The technology has already been used to create fake accounts, profiles, websites and even fake video or voice conversations.

The fight to remain one step ahead of fraudsters will never end, but fintechs are keeping pace. New companies, such as Mantis Analytics, are using LLM and machine learning techniques to identify suspicious activity and fake profiles and conversations.

Compliance

Regulators are keeping a close eye on the development of LLMs and are particularly concerned about ‘black box’ models that are not able to justify their decision-making. Regulators will want to be reassured that decisions made by AI algorithms are logical, fair and non-discriminatory. The quality of data is essential; inaccurate and poor data will result in poor decision-making.

How we can help

Our aim at Nortal is to create seamless customer experiences that are human, valuable, intelligent, and resilient. We specialize in digital transformation projects and creating solutions that have a meaningful impact, helping our clients to transform and future-proof their organizations by building world-changing solutions with the right technologies. LLMs and Nortal Tark are an exciting and significant step in that journey.

Nortal Tark solutions are built around four key principles:

• They are LLM agnostic – LLM can be plugged into any Tark solutions.
• Customers will retain complete control over their data. We will always take data privacy extremely seriously.
• Our Tark solutions can be deployed anywhere – at the client’s premises, on the cloud, or using a hybrid approach.
• Our solutions will deliver value at speed.

LLMs bring enormous potential to the financial services sector and as we continue to develop our Nortal Tark suite of solutions, our clients will soon see the benefits for themselves.