Klea Neza explores how AI is progressing, and the challenges facing its adoption in the asset servicing industry
In a fast-paced financial environment, the hunger for successful Artificial Intelligence (AI) methods is a continuous battle between the different businesses that compete in the asset servicing field.
With industry demands constantly changing, the methods used by asset servicers to manage their work must adapt to these transformations, but this may bring a query to light: why choose AI over traditional frameworks?
It can be argued that through traditional methods, companies have a higher chance of succeeding with less risk. These types of methods normally cost less and are considered safer to carry through.
Despite the risks, the merits of AI are slowly being recognised by many companies, allowing them to access more efficient and accurate systems in their strategies. Research from great.gov.uk shows that the number of AI companies in the UK has rapidly increased by over 600 per cent in the last decade.
The digital consulting and analytics company EXLsurveyed 64 senior figures in the UK and banking industry and reported that more than 90 per cent of leaders asserted that they were successfully using AI to better their decision-making and to enhance existing products and services. They also note AI would be used to increase revenue and moreover, reduce risk.
When specifically focusing on AI in risk management, AI is able to process large amounts of data quickly, as well as identify anomalies along with many other benefits that come with the ‘robot version’ of risk management. However, that is merely a glimpse into what AI is truly capable of.
But it seems as if some businesses may still feel hesitant towards the use of AI. According to Psychology Today, the fear of change is undoubtedly unavoidable in human nature as a result of our brains resisting uncertainty. This can even be seen through the adoption of AI.
So does the asset servicing world take a leap of faith by integrating modern AI into their process of work, or do they cling to traditional methods and stay within their comfort zone?
Blinkers off — gaining new perspectives
Through different perspectives, traditional methods in asset services can be seen as a way of maintaining stability, or as a safety net to fall upon for those who may not want to try more modern methods.
But if something already works, you may ask: why fix something that is not broken?
Of course it may not be a case of the item being ‘broken’, but rather of finding a version of the item that is more suited to the specific situation, and will benefit you further than the first.
Consider a vehicle with an old engine that carries out its main purpose which is to keep the vehicle running. In the short-term, you will be saving money by not buying a new one and you feel as if the engine you already have may be suited to your demands.
However, through investing in a new engine, you may find that it has more benefits than the first one, such as it being fuel-efficient and producing lower emissions. In the long term, you will be saving money on petrol and benefiting the environment.
Now let us apply this to AI and traditional methods.
In essence, although traditional methods in asset services may not necessarily be broken, AI has the capability to optimise and enhance processes that could lead to greater efficiency, innovation and accuracy.
Therefore, looking at this topic through a different lens shows the embrace of AI methods in asset management does not have to be defined as replacing traditional methods, but implementing better ways of work into new practices.
By removing the blinkers of conventional thinking, we are able to value the use of AI and learn more about how it can be integrated into different companies in order to maximise successes in the asset servicing realm.
Sell side
You may question, why, and how, do some sell side companies use AI?
Aaron Armstrong, partner and head of sales at Intellimation.AI, a fintech company that provides global market solutions, asset management solutions, and insurance solutions, highlights the importance of AI. He summarises that “AI is applied in many different ways”, but focusing on risk management, which is of paramount importance to a bank, intellimation.ai’s solutions “manage operational risk, financial risk and regulatory risk.”
Through AI, the companies are able to make processes “further automated, audible, consistent, secure, traceable and accurate”, he adds.
Despite there being a variety of technological improvements in the way investment banks manage information, the combination of legacy systems, an ever-increasing demand for data, and the fact most data is unstructured, leads to many ‘human in the loop’ scenarios. So by definition “human error” is still a factor, highlights Armstrong.
Traditional vs AI
Though traditional techniques can perform the same act, AI can report back to regulators faster, more effectively and with a higher degree of accuracy. Importantly, they can also extend as the regulators and require more information. Domain focused AI can cut through inefficiencies in various workflows, with its ability to contextually interpret structured and unstructured data and align with convention. These solutions are able to target industry-wide problems, email itself being one common source, as Armstrong describes.
He comments: “A big bank will receive thousands of emails a day from customers, asking about whether they should be levering or receiving collateral on the back of certain trades to offset the counterparty risk.” He describes this process as ‘non-standardised’, involving human free text and multimodal communications. He continues: “The big banks will have teams of people watching inboxes, having to read all these emails, including all their attachments (CSV files, PDF’s, spreadsheets and more). And then on the back of what’s being requested, they have to check their own internal systems to make sure that the request itself, the type of, and amount of collateral, is appropriate before taking an action on the basis of agree, disagree or partially agree.”
The process of manually checking these emails does not only take a long time, but can be tedious for employees, and therefore leading to even more human error. These errors may result in a loss of profits, as money is continuously put towards correcting them, in addition to the problem of customer satisfaction itself. Thus, in order to save time in the long run, AI is able to fully automate this process. With the use of vertical AI, Armstrong explains that 95 per cent of these emails will be processed to 100 per cent accuracy. That then leaves five per cent of emails to be checked by “humans in the loop to manage edge cases”, he states. In doing so, employees may be able to have more time to focus on higher-valued tasks, which is better for both the bank, and the employees as well.
Regardless of AI already aiding financial institutions positively in such a short period of time, Armstrong brings to light that “It’s just the start of this great journey.”
He illustrated that where AI advances processes, traditional methods gradually slow them down. The challenge that ‘old’ technology brings, is that the technology is based on human rules, and what was initially written as human software. Armstrong describes this problem as running into a “glass ceiling”. As the analogy suggests, old technology is an obstacle that a company is unable to see and thus, can pretend this systematic issue does not exist. As a result, instead of moving forward with innovation, some businesses choose to maintain old techniques.
Furthermore, traditional methods can include repetitive techniques such as robotic process automation. Armstrong questions: “if your process is all about repeating, then how does the system know right from wrong?” Which may leave us to ask, if the system does not know the difference between which is right and which is wrong, can it truly be accurate and thereby actually scale? Intellimation.ai prides itself in being change and volume agnostic.
AI limitations
Those who prefer traditional methods over AI rightly argue that AI has its weaknesses. As Armstrong previously mentioned, AI is at the start of its journey and there are many improvements to be made in the way companies in the financial industry carry out their work.
Armstrong notes that “not one size fits all”, hinting at the idea that companies must know what type of AI is best suited to them in order to be effective. This can be seen through Intellimation’s use of vertical AI which can be defined as a type of AI that is created to carry out tasks and adapted to its specific industry. Intellimation.ai’s models have been “trained on banking financial services, data, and that inherently raises their accuracy”, due to being task specific, which gives it a head start over both traditional methods and language models.
However, if the financial industry uses AI that is not specifically processed to carry out their tasks, they are at risk of using AI ineffectively, and therefore having a negative perception of these models that, if used correctly, would be much more efficient than traditional techniques.
Although Armstrong did not seem to mention many weaknesses to AI in the financial realm, he did hint at limitations lying towards generative AI in which he asks: “Would you use generative AI to automate?”
He continues: “Generative gives you that risk of hallucination, or at the very least their output is probabilistic, which is not ideal in the world of moving cash flows.”
Armstrong highlights that the problem people may have is that “they’re trying to apply large language models to every problem. But there’s other aspects to AI that are a better fit for purpose”, illustrating the reason why companies may be hesitant to use AI is because of their uncertainty on which type of AI is better suited for their needs and client demands.
If businesses were more educated on the different types of AI and models, there’s a possibility that they could progress much quicker than those who continue to use older technologies.
Intellimation.ai’s models and solutions only shed light on how effective AI can be in the financial industry, specifically banks. Armstrong considers the technology they have built highly “interoperable”, to be the glue between old and new systems, “data fragmentation” also impacts risk management negatively. He says: “You can have the same data in different silos with different names. And when you try to reconcile all that data people find it very hard because even though it’s the same data it’s pulled by different names.”
The process of Intellimation.ai’s model enables them to sit on the legacy systems, enhancing and augmenting them into the new AI paradigm. Essentially through the use of their AI models, they remove the problem of data fragmentation, highlighting how AI more generally, can be more effective than traditional methods.
Final input
To the companies that are hesitant to use AI — don’t be.
Armstrong mentions the idea that “you need to be cognizant of what’s the right application of AI for the right problem statement”, and if you succeed at this, then you have a much greater shot that your AI can be implemented smoothly into your current systems. He further advises companies to take ‘baby steps’ with AI, if not used in previous systems before. Working with someone who has expertise in this field and is able to advise you can “turn a leap of faith into something truly transformational,” he adds.
Although it is easier said than done, it is clear that client demands in the financial sector are increasing rapidly and in today’s age, old systems that only use traditional processes may slow a businesses growth and success rate. The comfort zone only exists as a result of a familiarity with certain processes a company is used to, however changing demands also means changing techniques to match these demands.
Implementing AI may seem like a risk, but the only risk a company may be taking is not trying something new, because if you do not at least attempt it, how will you know it will not succeed?
