Steps Ahead

How to become AI-ready: three foundational steps to success

In a recent survey of financial services firms, 80% confirmed they were already using AI in some form and 90% stated they saw it as giving them a competitive advantage for the future. But underneath the hood, is this new technology really being fed the fuel it needs to provide accurate and useful outputs for banks’ sales and trading teams?

The elephant in the room: data health

Banks are investing vast sums of money and resources on AI initiatives, spending $20.6 billion in 2023 alone. The aspirations are clear – AI can deliver numerous benefits, including improved productivity and efficiency, reduced costs, better risk management, enhanced customer experience and greater innovation. However evidence from a recent study suggests that little ROI is typically achieved by AI deployment in one of the most critical divisions of the investment bank: the capital markets front office.

So, what’s the problem? Nine times out of ten, the issue lies in failing to ensure the data being fed into AI tools is in a good state – aggregated, standardised and enriched – before embarking on a multi-million dollar initiative.

In the current economic environment, ROI remains front of mind for all banks considering investing in AI to improve the efficiency, productivity and profitability of their front office – and data health should be where their journey begins.

Step 1: Aggregation

A significant proportion of financial services institutions struggle with data fragmentation, with multiple data sources and reports being generated across the front office. This is particularly pertinent in markets such as FX, where there are a large number of electronic trading venues, or corporate bonds, which are still often traded via voice.

This fragmentation can produce inefficiencies and may lead to data inconsistencies across different reports or teams. On average, five to ten data sources, including transaction, market and static data, need to be joined to get the full picture of flows across the organisation.

The first step on the road to AI deployment must be aggregating these disparate sources into one holistic database, so banks can be sure they are using all the data they have at their disposal.

Step 2: Standardisation

For it to be truly valuable when AI is applied, the very fabric of a firm’s data must be integrated and used in a way that is frictionless. As such, once they have been aggregated, data sets must be standardised across asset classes into one consistent format and cover as wide a set of relevant transaction and market data as possible.

This is, however, often far from straightforward. Within the FICC markets, each trading network adheres to its own messaging language for passing and recording trades and there will often be wide variation in the data definitions and fields captured for a given trade. To add to the complexity, data which firms bring in from external sources will have been processed in a way which is unique to that data provider and cannot simply be added to this new unified data set.

Without data standardisation, the challenge participants face is that each trading channel taken in isolation provides only a partial impression of market activity. Basing AI initiatives on such partial and narrowly applicable information flows will, at best, lead to compromised outcomes and severely limit the value which can be derived.

Step 3: Enrichment

Once it has been aggregated and standardised, each data entry should be made as comprehensive as possible, with all relevant fields captured for every transaction.

This is where external data sets can be employed to ‘plug the gaps’ in a firm’s data. This includes enhancements such as using market data to enable market impact comparisons between the firm’s activity and the markets as a whole, but it can also include far more complex additions to the data set, such as introducing risk calculations onto the data record for cash or derivative trades. When a firm is considering its position for any instrument – spot, forwards, swaps, futures, and more – the need to really understand what is happening in the markets is an imperative.

Building on the foundations

Only once these three key data health foundations are in place can a bank begin its journey to successful AI implementation. Unfortunately, however, the vast majority of banks do not have their data in a place where this is even possible.

Critically, what this tells us is: banks must address the fundamentals of their data business before expanding onwards into further technological development. Once these steps are complete, banks will be able to capitalise on the power of AI to provide real-time and actionable intelligence at their fingertips.

By choosing to engage with a specialist data analytics provider, banks can benefit from the most relevant and up-to-date thinking around the marriage of transaction, market and reference data and in so doing, time to market for value added AI projects is shortened and, at the same time, at a significantly lower cost than an internal build.

Preparing big data for AI

Financial market infrastructure providers (FMIs) and other large financial institutions sit on a wealth of untapped data from the markets they facilitate, but typically struggle to convert this raw unstructured information into a format in which AI can be applied to provide true market colour.

Smart Markets from Mosaic Smart Data converts big data sets from FMIs such as Euroclear into a standardised format and contextualises it, applying cutting-edge AI-based analytics to enable market participants to enhance trading models and build informed strategies.