AI in Capital Markets: From Automation to Insight-Led Decision Making

AI in Capital Markets


AI is not just enhancing existing processes, but also bringing a larger spread of tech offerings like natural language processing, robotics and cognitive systems. These technologies are increasing revenue, enhancing customer satisfaction and bringing forward more business. The call of the hour is for BFSI stakeholders to actively engage in AI adoption and explore its past potential for industry transformation.

Stefano De Marzo | Head of News at EU-Startups


Stefano De Marzo

Stefano De Marzo | Head of News at EU-Startups

Anand Chandra

Anand Chandra | Growth Leader, BFS, EU and APAC at Accolite

Episode Transcript

Anand Chandra: Hello all this is Anand Chandra this side.

I welcome you all to the session, AI in Capital Markets. In today's session we would look at the “Spend-to-Save” approach and what are some of the themes that go around in financial services. Specifically in this case, in capital markets. Through this session we will try to bring what we see, what we hear through our customers through our prospective clients, some of the industry SMEs. And you would hear Stefano speak and those would be some of the themes that we would go through.

I will do a quick introduction. I'm one of the group leaders for Europe and APAC. I globally lead the Center of Excellence for BFS for Accolite. I'm joined today by Stefano. 

Stefano. I'll let you introduce yourself.

Stefano De Marzo: Good afternoon, everyone. I'm Stefano De Marzo. I'm head of News of EU Startups - a media outlet covering the European innovation ecosystem.

I've been a contributor for Forbes, Entrepreneur magazine, Sifted - now get backed by the Financial Times. And I've been covering for many years emerging tech, startups, innovations and venture capital.

Anand Chandra: Thank you, Stefano. And with that Stefano, it would be good if you could give us a view from a journalist lens, from an SME, who looks at tracks that keep a tight view on the pulse of what goes around in the market. So it would be awesome if you could share some of those findings.

Stefano De Marzo: Yes, thank you, Anand. Of course. Well, today, I have to say that we are in front of an exciting frontier. That is the evolution of artificial intelligence, and its profound impact on the banking, financial services and insurance sector.

This means that the journey of AI, from its inception to its current state has been nothing short of remarkable. And, for example, Gartner, the Global Research Advisory firm has recently positioned generative AI at the peak of inflated expectations on the 2023 hype cycle for American technologies.

This indicates a transformational benefit awaiting us in the near future.

When we talk about the AI hype,this means in a way that everybody's following the crowd, without paying too much attention to details, or to reality, sometimes. But, as we move beyond the hype, the potential of AI and revolutionizing customer services, by assisting in data processing, drafting engagements and qualitative - quantitative analysis is undeniable.

This is transforming communications between analysts, portfolio managers, and clients, who are sharing an era of automated investment processes and personalized portfolio management through robo advisors.

In short, this technology is enabling organizations to make better investment decisions. We have machine learning algorithms analyzing vast amounts of data to identify trends and predict future market movements, supporting better investment management, and decision-making. But, before going into some EU examples, we have to say that AI, particularly through robotic process automation, is reshaping banking operations by eliminating day-to-day use error-prone data, entry tasks and enhancing capabilities in handwriting recognition and natural language processing, we are witnessing a synergy that results in intelligent automation, significantly reducing operational costs and risks.

Furthermore, AI powered chatbots, and conversational assistance are now providing round the clock customer support significantly enhancing customer satisfaction and engagement.

For instance, the daily routine of bank employees often revolves around time consuming and repetitive tasks. Through intelligent automation, these tasks can be automated, delivering more accuracy and allowing employees to focus on priority tasks without distractions.

Right now, in the European Union, several banks have successfully integrated AI to enhance their operations,customer service and decision-making processes. For example, Danske Bank, Denmark's largest bank, has leveraged AI for fraud detection, implementing a fraud detection algorithm that significantly improves its fraud detection capability by 50% while reducing files positive by 60%.

Another example is Ulster Bank, one of the 4 biggest banks in Ireland. Ulster Bank, collaborated with Atos, a France-based company to enhance his customer experience. Atos assisted Ulster Bank in optimizing the utilization of salesforce new AI to Einstein.

By cleaning and organizing the bank's data to fit the format requirements for Einstein.

Atos developed an Einstein-based next best product recommendations engine for the bank's relationship managers, helping the bank to segment their customer base more accurately.

Another example is ING, the Dutch banking service provider. ING developed an internal AI tool called katana, to aid bond traders in making better buying and selling pricing decisions, using predictive analytics. During pilot testing it was revealed that katana enabled traders to quote prices faster than 90% of trades, reducing the cost of trading and allowing traders to quote the best possible price, more frequently. So, looking towards the future AI driven chatbots are set to provide personalized customer support addressing increase and concerns in real time. This automation of mundane tasks like balance increase and pass for research, presets will free up customer services representatives for more complex issues leading to cost savings, efficiency and therefore saving availability.

But what are the challenges? The BFSI sector often faces several challenges, such as incremental fragmentations for once. This leads to data silos and inefficiencies. Over time, these challenges compound, affecting communications and data sharing between departments.

In this case AI stands as a beacon of innovation, offering solutions that not only address these challenges, but also unlock a multi-billion dollar opportunity by fostering cross functional collaboration and reducing the hidden tax on innovation known as software entropy.

Another challenge is when it comes to the data intensive nature of AI, this calls for enhanced security measures to protect sensitive information of course.

The requirement for high quality stricter data necessitates adjustments in data policies to ensure privacy and compliance. Sometimes the lack of explainability in AI decision making requires us to strive for a better understanding, validation, and explanation of AI generated decisions to mitigate biases and maintain trust.

In conclusion, the transformative potential of AI in the BFSI sector, as predicted by Gartner in this graphic, is an indication to a renaissance of operational efficiency, customer satisfaction and innovative solutions. AI is not just enhancing existing processes, but also bringing a larger spread of tech offerings like natural language processing, robotics and cognitive systems. These technologies are increasing revenue, enhancing customer satisfaction and bringing forward more business. The call of the hour is for BFSI stakeholders to actively engage in AI adoption and explore its past potential for industry transformation.

Now I hand the mic to Anand to continue further with the webinar.

Anand Chandra: Thank you, Stefano. Thank you. It's a good perspective. If I look at the graph in terms of what the trigger points are, what the inceptions are, in terms of how the expectations are set. And then there is an element of disillusionment that has it in the way. Then there is an enlightenment aspect which is almost like an epiphany you would have in terms of what you see and how you visualize it. And then there is a productivity aspect that kicks in.

Very similar to that. If we were to look at it from the lens of an IT services organization like Accolite that specializes in digital prompt engineering, how do we look at it? Simplistically put across, opinions vary as it is clamor. But if you look at the different stages of how the industry is visualizing AI- you have natural intelligence, a narrow aspect of it, then you have general intelligence, which is where the AI comes at par with humans. It is an absolute same clone and then there is a super intelligence aspect, which is quite out there in the future in terms of 2050, 2060.
So here we are today, narrow intelligence plus plus plus, is where we as Accolite visualize. Now, you nicely described the progression in terms of robotic process automation, the mundane, mechanical, repetitive task, that is where the traditional manual work flows. If you had to have automation towards the complexity side of it, so, we right now see it somewhere in the middle, and that's where the green dotted line is from an AI driven innovation perspective. A step before that is what we have seen for roughly half decade or so, which is largely in terms of machine learning, machine learning models, predictive analytics in terms of what data curates. How do you interpret some of that data? The decisioning part has not necessarily kicked in very recently it has. You gave an example of ING from a bond trading perspective, quite an exhaustive and labour-ative task. If you have to price a corporate bond, or a convertible bond that has a lot of swaps attached to it. But if you have a bit of an AI built into it, that is where truly a born trader would be able to do it, leverage that off.

Anand Chandra: It has largely limited itself to a predictive side of it, and what we now see is actually execution of those trades well, in several forms and shapes. Some of the examples you gave are already part of those AI driven innovations. We're looking at it from a portfolio optimization, portfolio rationalization.

You spoke about fraud detection. Real time transaction monitoring.What used to be a T+0 end of day activity has now actually become quite real time, and that is where some of the conversations that we are having with our customers and prospects of Accolite, is inclining towards. Then you have credit scoring.

We'll give you a shade of what we are developing as an industry solution. But AI driven credit scoring is, I would not say it's not there, but is right now it is now, you know, almost in a self, learning, self-paced mode. Right? And then you have AI driven trade booking, which Stefano you described from an ING example where it is maturing.

Where does this headed General Intelligence, which is where AI and humans would come at par. Still, out there, up for debate, there is an ethical aspect of it.

There is an usability aspect of it. Then there is an element of how do you bring that balance between human and a machine aspect of it? But then that's where reasoning and problem solving would then be handed over to AI. And that would be, that would be a true AI state in terms of what we are looking at.

Strategic objectives for capital markets, financial services as a whole in terms of customer experience, Stefano, you touched upon chatbots, improving loan and credit decisioning process in terms of human intervention versus what machine is taking. Then you touched upon fraud detection, regulatory compliance. There is an element of automation in the investment process, altogether.

Identification trade booking, confirmation, settlement, and all the key plus one key plus 2 activities in terms of reconciliation and regulatory aspect of it. So those clearly are the strategic objectives that we see as an industry there. Now, the second part of the webinar, we wanted to share and showcase some of the industry solutions that the Center of Excellence is working on. Now these solutions are in today's interest of time. We are just showing you a couple of them. But it's an exhaustive list of roughly 40 plus use cases that has internal team brainstorming. We clearly have not ventured across all the 40, but we are looking at prioritizing them and bringing it to our customers and prospective clients in terms of what applied innovation could look like. Now, before I take a delve  into that, I'll just quickly summarize who we are, what we do, so that the audience has a context of what Accolite is and where do we fit in

Purely purely digital product engineering firm.

Roughly 3,000 people globally, operating across North America, Europe, we have a sizable presence in APAC, roughly, up to 35 clients of a full global 2,000 list. This is where we sit.

Since inception, product - owner role is ingrained. What the industry picked up a decade ago has been there to a core of what we do and how we operate, and how we deliver.

Anand Chandra: And that is where some of the conversations of Accolite are headed as well. So that's a quick brief of who we are, what we do. Now without taking too much time, I'll touch upon the two industry solutions.

At the end,the audience has an option to ask a question. Should you have any questions, we are happy to answer them as part of the webinar, and or please reach out to us very happy to do a, demo - sorts, or further, more detailed walk - through for you if any of this is of interest. Now, cash flow forecasting, it would be unfair to say that the function does not have automation, the function does not have AI; the aspect of the business in terms of treasury has not seen modernization.

I would be lying to you if I said that. Having said that the usage of the AI that has been has been incorporated, the modernization that has happened, what we now see is a lot of that modernization has developed a bit of duplicity in terms of different aspect of treasury functions have ventured upon their automation, and same feature sets have been automated across different application base. The AI that has been implemented to whatever degree, in some cases it is the machine learning models, is there, but it is still trapped in silos, which is visible, which is interacting and exchanging with a very particular aspect of a treasury function.

Cash flow forecasting as a whole is part of the liquidity payment manager. So through our framework we are able to showcase how that entire process of cash flow forecasting can be automated. On top of that, how do you apply optimization strategies? How do you visualize different scenarios?

Now those scenarios could be as simple as a liquidity coverage ratio. It could be a high gross profit margin ratio aspect of the business, and at the same time, how do you integrate compliance as one umbrella, almost like a straight through process for a treasury function. Now that is something that we have worked upon.

There is AI built into it. All the industry solutions are being developed on the no - code tag. Entire aspect of it is connectivity through API. So we are showcasing how APIs and micro services are front and center. We are not talking about it because it's there. There is nothing non API about it, It is exchanging, giving, interacting everything in the API base.

And last, but not least, while it has two AI models plugged into it in terms of cash flow forecasting, and some of the liquidity coverage ratio in terms of thresholds, that we would calculate. We also have a generative AI piece built into the liquidity payment manager, which is sitting on top of automated cash flow forecasting.

Now that generative AI piece is largely around, you have surplus cash.

You are short of cash. You will have a process that will do a crash sweep. It could be overnight. It could be real time. There are different ways and methods you can set it up.

The interesting part is, if you have surplus cash, how do you maximize it?

How do you take that surplus cash for that duration of surplus time that you have, and make an investment out of it.

That is where the generative part comes in. We have what we call a synthetic trade broker into the treasury function. It would book your currency hedges against the cash portfolio where traditionally, historically, based on the AI model, where your cash performance has been slightly fluid and that is where the trades would be recommended. They would be then booked and sent to the OMS. In terms of a new status, which is where still the middle office needs to execute the trade, so that entire process is through the application. And that is what we are trying to showcase in one of the industry solutions that we have.

Now, what business benefits we are delivering through the liquidity payment manager. Automated cash flow forecasting is the core engine that is within that framework is innovation as a service. I mentioned AI models. I mentioned the generative AI part of it to synthetic trades. The synthetic trades by the way, the element of the trade compares the different rates across fed funds, short term liquidity investment vehicles, that you have across different institutions and vehicles, but we have kept it to short term and medium term, because this is cash transactions. So you need to make sure that your treasury functions have enough cash. So all of that data, gathering, scraping, and then analyzing that data and based on that creating the synthetic trades is what we call the innovation as a service in terms of gen AI.

Clearly, operational efficiency is something that you are delivering in terms of reducing risk. Stefano touched upon manual intervention, duplicate task, repetitive task. And that is where the efficiency part comes in from an operations perspective.

Treasury API factory - You could use this as a standalone solution, you could use it in parts connect to already modernized system, and or something that you are venturing upon to get it modernized in terms of tech stack, in terms of connectivity, in terms of an API marketplace, and this industry solution would help you bootstrap that process. Liquidity optimization through automated trade booking I've already mentioned. Automated compliance, there is a risk management and compliance portfolio within the application. It talks about different edge cases. It talks about internal audit guidelines. It talks about some of the reports that are pending, to what degree those reports are able to service based on the data contracts that have been established, all of that is blended into a compliance module within the framework.

Last, but not least, not that we were leading with this. But it's a byproduct - lowering your total cost of operation is something that you would get, through the liquidity payment manager.
Now, the second one that I would want to cover is an Interactive Investment Recommender. Now think of a day of an analyst, of a portfolio manager or a wealth manager or an asset manager.

The analyst spends several personnel hours to the tune of 4 to 8 weeks looking at the available list of prospective customers, looking at some of the new products that the organization is packaging, curating, or has a tie up with it could actually offer it to your end customer as a portfolio manager or a wealth advisor.

And then the entire effort goes in terms of comparing List A with list B and how do you bring this all together. There is a decent amount of automation industry. Again, I would be incorrect in saying that there is no automation. There is no AI, but to what degree it is able to make decisions is the core aspect of it. The automation is largely about reading all the documentation and creating one pdf which would be one big monolith pdf which would talk about several customers and several products that they can map in terms of a permutation on a combination. But it's not interactive. It is still static. What we have created as an industry solution as part of our Gen AI, Center of Excellence is a very simplistic AI set of models and then there is a generative part that sits at the back end.

You log in. There are 2 personas. You have a product list. You have a customer list.

You click the customer. You select one of the customers. It has a pre assigned risk code, and when I say pre assigned, it is calculated by the solution based on several factors, investment appetite, portfolios split across equity commodities, fixed income exotics, and several other asset classes and the demographic details.

So it talks about persona. It talks about investment appetite, calculates a risk core, the AI model that generates it, is configurable. You can actually feed in different parameters to calculate on, the guidance based on which there is code can be generated.

The same thing that you would then have a list called as Generate Product Recommendation. You click that button.

The AI back end uses some of the Gen AI principles of chatgpt3.5. It produces an output, Mind you, all of this information is now getting curated, based on market sentiments, different data providers, different connectivity of an API.

This includes a degree of connecting to Bloomberg to get some of the information feed. So we have all those plugins built in. We have not necessarily connected all of it. But that engine is there.

And then scrape through and generate a product recommendation list based on what currently the organization hosts. Not only that, that's where the regulatory part comes in is where, if you were to recommend this product, and if the end customer is to execute that product, what that recalculated, reassessed risk code would look like.

Now that's the regulatory part. This is where the fair aspect of it, transparency aspect of it comes in, and all of this is under 3 or 4 clicks, backed by AI and Gen AI. That sits on top of it. Again, what do we deliver as a business benefit,  innovation as a service, talks about approach, risk or patterns, market risk, data scraping. That is where all collected together.

Customer 360 - You could look at customer and indicative recommended product lists, or you could go to a product and generate a set of customers that would be best suited for that product.

Very simply put, it's a marriage or a match making of a customer and a product recommendation, but very interactive in terms of 2 or 3 clicks that would bring both together.

Chat with your data - it is the heart, it is the core.

I mentioned data avenues, data, chapters, data nodes from where the data would come through APIs into the system. Again, the platform is built on no code.

All the industry solutions that we are developing, most of it from a BFS, capital markets perspective are getting developed on no code. Not a single line of code is being written, and all the business functions are still packaged as a modular and an API that can be extracted and plugged in and played across different aspects of the trade life cycle.

That is what Accolite is looking for from an industry solutions perspective at this point in time. I touched upon integrated compliance. I touched upon risk or reassessment. That's the regulatory part of it. It's keeping the customer at the core when you actually do your regulatory aspect of the reporting, and not necessarily the institution that is sending those reports and that's the fundamental shift that you would actually see as the evolution of AI goes from narrow intelligence to general intelligence, where the end consumer or the customer has to be key, and it should be understandable, it should be transparent. All the AI that you use has to be explainable. It cannot just be a machine that sits and does something, and you have to believe in it. That is what Accolite is seeing in terms of some of them that are coming in. 

Cost reduction, as I said, it's a by-product. We are not developing these industry solutions necessarily with the view of cost reduction. It comes through the process of optimization, efficiency, gain, and so on. And those are some of the themes I wanted to cover. Now with that I'll come to a conclusion in terms of the presentation that I wanted to give. Like, I said, these industry solutions are not exhaustive. We have roughly a catalog of 40 plus use cases that we are working on. It is still in exploratory mode. We are still interviewing some of our CXOs and industry leaders, and seeing if some of the new themes can be included into our Center of Excellence, where we can build these solutions and take it to our customer. The IP is still with the customer, we are pure services that way. So I'll take a pause. I'll check with Stefano if he has any questions. But that's largely the part I wanted to cover. Stefano, anything to add any questions.

Stefano De Marzo: Yes, Anand, one question I wanted to ask you is, how much of a game changer can this be for a company to be an early adopter of AI related technologies.

Anand Chandra: Stefano, you gave examples of ING that you touched upon some of the use cases that are looking from a Europe perspective acolyte as a global organization, we interact with some of the North America customers and Asia customers as well, we see the innovation curve slightly different across.

But to stay ahead of the curve, why you have to adopt an AI now is best used for human capital.

There is this view that AI and automation is having an impact in terms of, you know, there is a lot of reusability of people's time and effort. And that is where the core business of a bank can actually fully use the time of the people on focusing the core business side of it, servicing the customer and leaving all the innovation partners like us and to AI as a technology, that would then allow that automation.

Right now there is a lot of IT that sits in organizations like a large bank that's not doing IT. And that is where the use of AI can actually generate a lot. Then there is an element of how do you stay competitive? How do you look at it from a new mind share, a new worlded share. And how fast can you read it?

What used to be Fintech versus Banks is now Fintech and Banks, and that is where AI is also coming into the picture. So contemporary views and traditional views are getting together. And that's where AI adoption is very critical for the Capital markets or financial services largely there.

Stefano De Marzo: So, if you have to refer, if you had to recommend 2 or 3 things, a CTO should have in mind before embarking on AI initiatives in his company. What will those 2 or 3 things be?

Anand Chandra: Absolutely so. There is a laundry list of things that a CTO would have, but if I have to summarize from a 2 or 3 perspective, he would look at what would be the core engine that you want to do in a way where that function is reusable across the different aspects of the business. You don't want to do AI for the sake of AI. You want to have it as much usability as you can. One of the themes that a CTO should look at is or is looking at is cross domain financial inclusions.

Gone are the days where it used to be a very domain-centric play for a bank, saying, this is what a bank does. Now financial institutions, through this API marketplace are connecting to manufacturing retail, logistics to payments they are connecting to different aviation aspects. 

So insurance is a big play and that is where cross domain, financial integration, or intelligence. And then applying AI on top of it is something that a CTO would look for.

Last, but not least, a CEO should look at it - what would it mean for an organization in short term, medium term and long term, as opposed to a pure cost saving play. So that is where we are also seeing that there are some AI automation picks in.

But it's not necessarily elaborating into an innovation, because it's very costly. So that's probably one of the things that I would look at.

Now, Stephano, while I answer that one thing pops up in my mind. you get to speak different avenues, different places interact with these C- level folks.

How do you see the broad brush canvas of Innovation in AI and capital markets across Europe, APAC, maybe North America. I don't know how much of that exposure you have. But how do you see that regionally shaping it closely for an organization?

Stefano De Marzo: Well, I for sure I see many differences between APAC and Europe, I mean, first of all, we have to consider something. APAC and the EU are different regions for once, because the EU is a homogenous block. You know, there's the EU Commission, etcetera. And the APAC region, are basically different countries with different experiences.

So most of the things I mentioned at the beginning of the presentation, they are right now happening in the APAC region. We can see that AI is reshaped in the financial industry there like it is extensively utilized for customer service. There are AI chatbots, streamlining operations and enhancing customer experience. There are robo advisors, like in those on site these Fintechs that are gaining traction, offering tailor investment advice to retail investors.

AI is also being used in the landing process, which is particularly significant.

So we see a lot of activity in the APAC region. In Europe the situation is different, like the EU is preparing to implement the comprehensive AI act focusing on controlling high risk AI applications. So, there are different situations in different countries in the EU. Because right now, there's not a specific regulatory framework. For example, in countries like Germany and Austria. They are being a little bit cautious, very German. But France, for example, is actively encouraging AI using, combining fraud and money laundering.

So that's an interesting experience to have to consider. Luxembourg, for example, which is as we all know, a financial center is actively also promoting AI initiatives through collaborations between the government and private sectors.

Also, we can mention Spain, which, right now, the government is trying to prioritize AI through a national artificial intelligence strategy.

So, well, what I can suggest is like, we have to keep in mind what's going to happen with that AI act in the  PM Commission  in the next few months, probably next year.

And we can see what's going to happen. As we mentioned at the beginning of the presentation. There's already things going on with ING, Danske Bank. So there are many countries and many banks with interesting things.

Anand Chandra: Oh, fantastic! You're leading through the regulatory lands, the early adopters of Basel 3, and that it continues to be the same. The Asian market is still very innovation driven. First try it out and then see, and then build an innovation, a regulatory framework around it. I think that is where the thought process is. No. Very true, Stefano. Thank you so much for sharing the desk, sharing the mic. I really appreciate as an industry journalist coming in and speaking on a forum with us. I really enjoyed my conversation with you.

And thank you everyone for dialing in wherever you are dialing in from thank you so much for your time. The recording is available on Youtube, on our channel, you could definitely drive and see. Thank you so much. And have a fabulous day ahead. Everyone.

Thank you.

Request a consultation.

Set up a 30-minute consultation