The Opportunities and Challenges for Artificial Intelligence in Financial Services

Recent advances in Artificial Intelligence are creating huge opportunities for process improvement in the Financial Sector but do have some challenges.

Introduction/ Background 

The financial services industry in Australia has not been without it’s challenges in the last few years.  Downward moving real-estate and investment markets in Australia, increased competition from global players and the recent high degree of inspection and intervention from Australian regulators has forced Financial Services companies to on the one hand focus more on customer service and doing the “right thing’ for their customers. Whilst on the other hand this same environment is pushing these same companies to deploy new technologies to improve costs and to increase efficiencies to offset the additional burdens that new regulation is bringing. 

In an era of big data with large volumes of transactions the deployment of AI represents an opportunity for Financial Services organisations to improve their operations through increased cost efficiencies whilst at the same time providing better services to their customers.  The opportunity for AI in finance is so great that Gartner forecasts that 27% of Financial Departments will deploy some form of AI by 2020. Businesses that successfully deploy AI could improve profitability by an average of 38% by 2035 according to a recent Accenture report.

But what is Artificial Intelligence (AI)? And what is it not?

AI is a broad set of technologies which enables machines to reason, understand and interact like humans. These technologies allow machines to learn and form conclusions with imperfect data, interpret the meaning of text, voice and images and to interact with people in natural ways. As a result users must understand that the term AI is applied to a broad set of technologies from Conversational Agents, Machine Learning, Machine Vision, to Text Analytics to Natural Language processing. Each set of AI tools and techniques are unique with different expertise required for each.

When discussing AI it is also important to understand the difference between Narrow vs. General AI. Narrow AI is Artificial Intelligence used for a specific business process (e.g. recognising cancerous growths on MRI images or identifying fraudulent credit card transactions). General AI is where intelligence is applied across a broad spectrum of situations, examples of which might be C3PO from Star Ware or Data from Star Trek. Just like the examples used it is important to know that Narrow AI is very much here and now where as general AI is at this point science fiction probably another 30 years off before becoming a practical reality. It's important to define these limitations when talking about AI in any application.

Another factor to consider about AI is the concepts of personalisation vs. standardisation. Some people think that AI is about standardisation of responses or treating everyone the same. This is more a feature use of Robotic Process Automation (RPA) which is the automation of known and fixed business rules. The real benefit of AI is the ability to provide a personalised experience by incorporating a multitude of inputs and variables into how an AI handles and responds to its environments and to human queries. An example of this is a chat bot or conversational agent which automatically uses a customer’s previous transaction history and understanding of the person’s job function to present a series of menu options to a user specifically tailored to them.   

What is Machine Learning (ML)? 

One form of Artificial Intelligence is Machines Learning (ML). ML refers to mathematical models based on historical information that are constructed to predict the impact a specific action the machine takes will have on its environment. ML encapsulates a large set of mathematical modelling techniques which includes amongst other techniques the deep learning used in neural networks. The term neural networks denotes that the mathematical models are designed to represent in a coarse sense the function of a human neuron or brain cell. The term deep learning or Deep Neural Network (DNN) implies that the models created using neural networks are highly complex using a number of hidden layers to represent various aspects of a system.  

Why now for Artificial Intelligence and Machine Learning in Financial Services?

Artificial Intelligence has been around for a number of years. But it has increased in popularity in the last few years. So much so that Gartner (2017) now boldly predicts that "By 2020, 85% of CIOs will be piloting AI programs". But why the recent popularity? There are a few reasons:

First, we have reached a tipping point in the amount of available data. AI is heavily reliant on the volume, velocity and variety of available data. As each of these factors increases the better AI gets at predicting an outcome and making a recommendation based on that outcome. Whereas humans get worst at making decisions. It is well known that humans employ a series of heuristics or mental short cuts to make decisions when confronted with large volumes of data. In additional when given a choice between making the best decision vs. the quickest decision humans will opt for the quickest option. Finally humans are often subject to fatigue and challenges with cognitive overload when they have to manage many issues at the same time. This can often lead to sub-optimal decision making. Whereas AIs don’t have these same challenges. Thus with the advent of the Internet and now the Internet of Things the amount available data is growing exponentially and the need and ability for machines to make decisions is increasing with it as well.

Second, with the advent of cloud computing the availability of large amounts of low cost computing now makes AI and ML more affordable. AI traditionally requires a large amount of computing power over a short period of time to train the mathematical models used in AI and ML. Now with cloud computing offering the ability to purchase extremely large amounts of computing power, cost effectively for a short periods of time, AI becomes more practical.

The third and final reason for AI's and MLs recent surge into the limelight is the advent of better math and better tools to deliver the math. Recently improvements in the mathematical models as demonstrated in some well-known global competitions like ImageNet mean that AI is getting more accurate in its approximation of reality. Answers from bots are more human like and AI models more robust at handling a variety of inputs.

Examples of Artificial Intelligence in Financial Services 

Artificial Intelligence has been in use in Financial Services for a number of years now but is increasing exponentially. ML routines are used to build complex customer profiles with up to hundreds of inputs both specific to the customer and market based to proactively identify higher risk loans for Credit/ Risk Management in order to Reduce Revenue Loss. For example Alipay who have close to 700 million active users have been able to use AI to reduce their fraud loss rate to less than $5 to $10Million USD per year.

Chatbots or Conversational Agents (another form of Artificial Intelligence) are used in a variety of sectors to manage financial services enquiries. For example Sydney University recently deployed a Chatbot dubbed Finbot to manage the enquiries from it’s internal staff regarding purchase orders, invoices, enquiries about suppliers and expenses. Available 24/7 and with no requirement to wait for a telephone agent, the conversational agent improves the perception of service and reduces the number of in-bound queries that the University of Sydney’s finance department has to handle with their call centre agents. The tool also recognises the person making the request through already deployed Active Directory tools to ensure they only get access to the information they should get access to. 

Machine Vision, another form of AI/ML which uses computers to recognise patterns in photographs and drawings is also finding broad application.  For example insurance companies are using AI to validate property features when assessing premiums for insurance claims and are also using Machine Vision in the verification of large volumes of documents to identify potential fraudulent document submissions.

As Financial Services organisation grow bigger faster they are looking for more ways to automate common everyday manual transactions releasing their workers to focus on more value added activities. AI and ML provide a number of different options in this regard. 

Ethical AI and the Challenges with the use of Artificial Intelligence in Financial Services 

But there have been recent challenges with using Artificial Intelligence. The GDPR legislation released in Europe in 2018 in an effort to increase personal privacy is having global effects for vendors not just for their operations in Europe. Buried in that legislation were regulations which governed the use of data for AI and the use of AI itself.  AI tools are banned from using certain types of data to make predictions about an individual (for example using race to determine whether they might represent a credit risk) and the individual has the right to know why a decision was made about them which may prevent the use of certain types of AI in certain situations in the first place.  These regulations were brought in due to concerns about the un-ethical nature of some of the decisions being made by AI in the Financial Services industry in Europe.

There are also concerns about the issue of bias in AI decision making. This is where the developers of the AI select data sets to train these AI/ ML tools which are unknowingly impacted by the human designers own unintentional bias. Systems integrators and consultants who deploy AI/ ML tools in Financial Services and who do not understand this concept of bias can end up building tools which are not only inefficient in making decisions but also possibly unethical in the decisions that they make as well. 

The conclusion here is that financial services organisations are well advised to select experienced systems integrators and consultants to deploy these tools who are trained in the concept of “Ethical AI”, understand the legislative and regulatory requirements imposed by the GDPR and other regulations around AI and who will protect your brand when deploying these types of solutions.  

Deploying AI in a Financial Services Setting

All Financial Services organisations undergo a typical path when deploying AI solutions. Initial awareness leads to localised proof of concepts for specific use cases which then leads into local operational use. Eventually the organisation starts deploying the AI tools more broadly and will become fully mature as a user of Artificial Intelligence techniques.  

When moving down the AI Maturity Curve organisations have a choice of being broad in the types of technologies they are use or to pick specific AI technologies to deploy. Remember that AI includes not only Conversational Agents but also Machine Learning, Machine Vision, Natural Language processing and many other technologies. All these different forms of AI have their own mathematical models and different tools/ user interfaces for their deployment and use. Financial services organisations need to choose whether they want to develop their maturity in only a select few types of AI or whether they want to explore all forms of AI. Finally organisations need to understand where they are on the AI maturity curve, where they want to get to and what technologies do they want to use to become a fully mature AI organisation.  

Conclusions regarding Artificial Intelligence in Financial Services

AI and ML represents a huge opportunity for Financial Services Firms to improve their operational processes, increase efficiencies and provide for better customer service by providing a more tailored 24/7 readily available customer service experience. 

The challenge is that Financial Services organisations deploying AI need to pick systems integrators and consultants well versed in the issues of ethical AI and the regulations governing the use of AI to protect their brand and to make sure they are making the right decisions when using these technologies.  

David Goad has 20 years of consulting and business strategy experience and is currently a Postgraduate Fellow at the University of Sydney studying IoT and AI. He also advises enterprises on their AI, IoT and Technology Strategies and helps them make sense of a complex technology environments. If you would like more information on AI or IoT Strategy feel free to contact David at [email protected] .