Artificial intelligence is here to stay. And to change our lives. And it has developed at such a dizzying speed that it is difficult, more often than not, to keep up with it. This is why it can be said that in mid-June, financial services scored a goal in this race when Cecabank's project was admitted to the CNMV's sandbox: its AI-based financial innovation initiative is the only one of its kind in this controlled testing space for the digital transformation of the financial system.
But Cecabank is not alone. The benchmark custodian bank for Securities Services in Spain and Portugal is working with GPTAdvisor, a fintech founded in 2023. The goal of this association is to explore the advantages and potential of Generative Artificial Intelligence (GAI) in its application to investment services.
The initial project in the testing phase focuses on two use cases: improve financial literacy and education and transform the way an entity's product information is searched for and analysed. Generative Artificial Intelligence (GAI) in its application to investment services. For the CEO and founder of GPTAdvisor, Salvador Mas -recent guest in our Finect Talks-, the CNMV sandbox 'is an ideal environment for carrying out projects on Generative AI, which is a very new and very powerful technology, the adaptation of which to the world of consultancy services needs to be done properly'.
These are still in their infancy, in their early stages, but the plan already includes moving forward in parallel with additional use cases as needs and progress are identified. Aurora Cuadros, Corporate Director of the Securities Services Division at Cecabank, highlights that this experience allows 'bringing technological innovations closer to investment services' with a "comprehensive" perspective, combining technological and business dimensions with the regulatory framework.
Cecabank and Salvador Mas had collaborated before, as a result of common interests in technological innovation, so this current project arose 'in a natural and spontaneous manner', according to the bank's executive. The synergy existing between the two companies, focused on the B2B model and with a shared interest in improving their services, has been the key to their participation in the CNMV's 'testing ground'. We spoke to both parties to find out more about the ongoing project.
They have commenced testing the two initial modules, product information and financial education: What results do you expect to achieve, and what is your timeframe?
Aurora Cuadros: Yes. The financial education application has been implemented and we have already conducted a number of relevant tests that allow us to draw valuable conclusions regarding the results obtained. It is a conversational, natural language application that consists of a chat where users enter questions and receive answers. This chat is specifically configured to answer questions relating to financial literacy, and the answers are mainly based on the content of the CNMV's educational guides.
The level of responses is astonishing. The application is able to interpret the very specific meaning of the questions, and even the tone, producing answers that are very appropriate to the specific question: without adding incorrect information, combining the multiple sources of information available for the questions, applying reasoning, synthesising a large amount of information, performing brilliant comparative analyses... These tests invite us to be very optimistic about the potential of this technology, which is clearly demonstrating a high level of reliability. The second use case, relating to product information, has been implemented and will be tested shortly.
We expect to finalise the testing of these two use cases in early October.
Salvador Mas: We already have very favourable conclusions in these two modules. With our GPTAdvisor solution, integrated with CNMV training content and Cecabank's product catalogue, it is undoubtedly a very powerful solution, firstly for advisers and subsequently for customers.
It is important to control the sources and the data, understanding this is essential. The greatest risk of Generative AI is the credibility of what it says. The public ChatGPT, which we all use, is not a reliable application to provide financial information, as there is no control over the sources. Entities will need to develop their own Generative AI systems with their own data, content and business rules, under a controlled environment. 'The level of accuracy in product information must be "mathematical", as it can directly influence investment decisions'
Are these tools intended only for advisers and entities or are they also intended to reach the end user?
Aurora Cuadros: It depends. There is no single answer. It depends on the use case under consideration, the objectives of the implementing entity and the degree of reliability that can be achieved in each case. Our goal is to reach the end user, but for prudence's sake we may need to take intermediate steps beforehand, and certain features may only be available to fund managers initially, and then open to the public.
The reasons are logical: we need the entities implementing these applications to become familiar with them. Using the two use cases described above as an example, it is worth mentioning that they are very different from each other, both technically and in terms of the potential impact on the end customer. The level of accuracy in product information must be "mathematical", as it can directly influence the investment decisions of end customers.
With regard to financial education, it fits perfectly with the way this technology is used and the impact on investment decisions is not so direct.
That said, we would open to the general public the application for improving financial education before the product information application.
Salvador Mas: It is inevitable that they will reach the end customer, and in fact we could say that they have already, because anyone can ask ChatGPT about anything. What entities need to do in order to use this powerful communication tool is to put in place the conditions for it to be reliable for their customers, because the cost of making a mistake is high for a company, much higher than for ChatGPT, from which the user does not demand reliability.
'Entities need to put in place the conditions for it to be reliable for their customers, because the cost of making a mistake is high'
Aurora Cuadros, Corporate Director of the Securities Services Division at Cecabank. Image: Cecabank
And how are they expected to improve financial services?
Aurora Cuadros: In the long term, we expect this technology to help make investment services easier for end customers, more convenient, of higher quality and also more efficient.
A recent EFPA survey concluded that while 67% of financial advisers believe that AI will make their work easier, only 22% use it. What barriers do you think you will encounter in implementing these tools and how do you plan to address them?
Aurora Cuadros: As we see it, there are three main barriers that entities within our sector will have to overcome on their way to deploying Generative AI applications: (1) data collection and management, ensuring data quality, (2) change management within the organisation, and (3) the ever-present regulation.
Entities will have to increase their efforts in terms of data, at the organisational and process level, and even in the search for and recruitment of talent through the hiring of data scientists. On the other hand, it will be necessary to adequately plan, promote and manage the paradigm shift that the use of this type of tools may entail in everyday practice.
Finally, we believe that regulation is necessary and to date we consider it adequate. Clearly, entities will need to have an appropriate strategy in place to adapt to the prevailing regulation at all times, and it will also be advantageous to maintain a good working relationship with the relevant regulators.
Salvador Mas: With regard to financial advice, quality is more important than quantity of use. It is better to wait for robust features and a good understanding of the technology than to rush into something simply because it is fashionable but without understanding its risks. I already see many instances of the use of AI in the sector that are not very edifying.
What are the main advantages Cecabank expects to gain by implementing Generative Artificial Intelligence in its investment services?
Aurora Cuadros: The advantages we expect to gain from the implementation of AI applications are mainly an increase in the value of our services and greater efficiency.
We want to continue to live up to our customers' expectations and we know that artificial intelligence is something that the industry has an increasing appetite for and that, potentially, in the medium term it will be essential for the investment services we provide to remain competitive.
On the other hand, we see artificial intelligence as a great opportunity to review our internal processes and detect possible improvements that will allow us to achieve greater efficiency, ensuring the quality of our services.
