The ﬁnancial sector turned long ago to technology and to AI - and with good reason. Finance’s adoption of new technology was motivated by the sheer complexity of the data it uses to predict and analyse trends. In simpler times, we used open outcry trading pits to set prices.
Today, trades can be booked through electronic platforms and a system called Algo Trading helps automate a large volume of trades. Finally, deep learning algorithms are widely used to analyse trends, to spot trading patterns and to understand market movements.
Alongside these process developments, the ﬁnancial industry itself saw the introduction of new ﬁnancial products. These products themselves require complex pricing and risk calculation models based on market data. This data can include: risk factors, FX rates, bonds prices, yield curves and equity prices. With the increase of customers, banking saw the growth of static data (employee, customer, counterpart and company data) too. In short, the ﬁnancial industry is an incredibly data intensive one.
Integration of AI in ﬁnance
In this digital era, personal data is used across all industries. Google, Facebook, and Amazon have been the front runners in automating their processes and incorporating AI into their business models. Algorithms have helped them get a better understanding of their clients.
Similarly, retail banks needed to incorporate these methodologies. For example, it is common for them to use data mining models in order to understand their customer pool and to grow their business.
On the ﬁnancial market data side, hedge funds and asset management houses have been using algorithms, for example, to send large execution orders. Finally, machine learning and deep learning algorithms are used to forecast trends and patterns of activities, for example, by clustering data related to trading activity.
Another aspect of technology, which is becoming prominent in the industry is RPA or Robotics Process Automation. RPA’s ﬁrst role was to replace repetitive tasks and to help companies move away from tacit knowledge and more towards a robust technical architecture.
Using RPA helps companies optimise their internal processes. New regulations such as BCBS239 (Basel Committee on Banking Supervision), MiFID II and GDPR require banks to guarantee a full understanding and transparency about their data processing.
They must guarantee the security of the data too. Thinking speciﬁcally about GDPR (see box below), AI could actually provide solutions for understanding the data and generating the artefacts required by these regulations. However, one of the industry’s challenges will be to ensure that their technical architecture can accommodate the integration of AI processes. These new systems must be incorporated efficiently into any existing internal technical framework.
Over the last 150 years, our industry has seen many changes. These range from the nature of the products traded to the implementation of new regulations, and, of course, the increase in trading activity. The insertion of technology has helped the industry grow signiﬁcantly.
Undoubtedly, our journey into the digital era will require us to protect, understand and be more efficient with our data processing. As these changes gather pace we may, however, need to ask ourselves a key question: could deep learning models of ﬁnancial data impact the macro-economic landscape?
In this article, I focused mainly on AI in the ﬁnancial sector. AI won’t, however, be the only innovation that may have an impact on this landscape. Cryptocurrencies and blockchain are already dominating thought and discussion.
Within the AI ﬁeld, development in cognitive science, behavioural analysis and Neural Linguistic Programming (NLP) are becoming more prominent. In the last 30 years, the ﬁnancial sector has evolved from traditional trading pits into models determining trading trends. AI will undoubtedly be more integrated into the ﬁnancial sector. The key consideration will be to incorporate the right innovation to respond to the bank’s needs and help it to support its customers.
Finance and GDPR
On a daily basis, words such as big data, GDPR and AI are mentioned endlessly in the news. Data is now a key component for companies’ marketing, logistics, decision making and processes. Simultaneously, AI is becoming a crucial component of any technical architecture.
Alongside the increase in new digital technologies, the General Data Protection Regulation - or GDPR - is considered a step towards the ‘digital single market’. Here is some information that will provide an understanding of the data regulation evolution into GDPR, and its impact on RPA (Robotics Process Automation) and AI.
- Evolution of data regulation
From 25 May 2018, companies using or storing data related to EU citizens, regardless of their location and where the data is processed, must comply with the EU regulation, known as GDPR. They aim to guarantee the Article 8 ‘Protection of the Data’ of the EU Charter of Fundamental Rights.
- GDPR key principles
As speciﬁed in Article 7, the goal is to build ‘a strong data protection framework’, by implementing the following:
Change in territorial scope: Protect data privacy for all EU citizens, no matter the location of the company.
Interaction with international organisations: Chapter V covers rules related to data usage by non-EU organisations.
Fines: Chapter VIII, Article 83, covers the administrative ﬁnes. These fall into two categories: up to 10 Million Euros, or up to two per cent of the total worldwide annual turnover of the previous ﬁnancial year; and up to 20 Million Euros, or up to four per cent of the total worldwide annual turnover of the previous ﬁnancial year.
Clear consent: The consent must be clearly stated (Article 32).
EDPB: ‘Article 29 of Working Party’ will be replaced by the European Data Protection Board (EDPB).
Rights of data subject: Right to access; Right to be forgotten (right to erasure): Article 66; Protection by design: Article 78
Breach notiﬁcation: The supervisory body should be notiﬁed within 72 hours of the breach (Article 85).
New roles: New roles including a data protection officer: Articles 24 to 29, 31 to 39.
Record of processing activities: (Article 30): Keeping a record of processing activities.
- AI and GDPR
In the design of AI algorithms, using aliases, encrypting and pseudonymising the personal data, will help us to comply to ‘protection by design’. For example, in Article 34 and 35, GDPR covers speciﬁcally how DNA (Deoxyribonucleic acid) and RNA (ribonucleic acid) should be considered as private data, and should not be linked to a physical person.
Section 4 of GDPR covers the ‘right object and automated individual decision making’, and requires transparency speciﬁcally regarding the proﬁling of users. On the other hand, certain AI methodologies, such as data mining, could actually help companies to search for the necessary information and generate the artefacts required by the regulation.
What is next?
One of this year’s World Economic Forum discussions was on the development of AI and the guarantee of the protection of private data, hence it is key for companies to be ready. If you feel you are not prepared for GDPR:
- Evaluate the extent of private data usage in your technical architecture.
- Check ICO (Information Commission Office), checklist for organisations.
Ginni Rometty, CEO of IBM, recently mentioned that by 2020, the private data value will be around one trillion Euros. By providing transparency on data usage, undoubtedly you will build your customers’ trust, and improve transparency in your data processing and algorithms.
Will GDPR be the step towards guaranteeing transparency on data and algorithms? Only the future will tell us.