Dan Ellis, CCO, Lumilinks
It’s impossible to ignore the rapid interest in AI development, or how it is set to transform any number of industry sectors. Deloitte recently found that well over half of UK workers have already used it, while within finance, they also found that over two-thirds (67%) of CFOs believe that AI will help to raise productivity. However, for AI to truly deliver cost, time, productivity and other efficiencies within the financial sector, it needs to be drawing on and learning from a robust, clean, and solid layer of business-wide data. Here’s why financial businesses need to first focus on getting their data houses in order, to capitalise on the AI revolution.
Unstructured data assessment. Finance moves quickly, and for AI or machine learning to make a difference, it needs to be working with the most recent , accurate and reliable insights. Many financial businesses still do not have all of this information in a single unified structure, and instead can even operate in very clearly defined silos. In fact, recent research has found that 9 in 10 financial services businesses in the UK and Europe are currently held back by heritage data silos. For an AI to deliver and to really get the most from the machine learning promise, first a widespread data audit is essential to understand the volumes of structured or unstructured data, and to close any gaps. We quite often start with businesses by completing a Data Fitness Audit which clarifies for businesses the current state of play regarding their information. This will also identify which systems may be set up to play nicely together, and which may need additional work before any AI plans to integrate.
Expectation management and project development. A CFO may have an ideal picture of a wonderful, AI-fused dashboard which can tell him at any given time the precise state of the business and allow him to interrogate at will. If they are looking to AI to drive this, and the data has not been assessed and audited first, it can at best depict only a partial picture. Does this mean that AI is better suited to individual or limited, smaller projects within financial services? Not necessarily. However, working towards a seamless AI-driven dashboard is a process which may take time, and advice, to deliver.
Financial services are nothing if not cautious when it comes to making changes within a business which might have regulatory, compliance or even customer privacy implications, and this may take more time to deliver than a turnkey AI-efficiency driver. This means looking at current compliance suitability alongside data quality and structure, and recognising that the more ambitious the automation, the more time it may take to deliver. Working with financial business among other industries, all too often we have seen promising data projects get dropped due to pressing priorities or project expectations outstripping capabilities. Yet, when we can see a project through to the end properly, we can guarantee a 100% success rate.
Dealing with the data elephant in the room and preparing for future scale. Studies have found that 85% of data-led and AI projects will fail. Venturebeat reported at 87% of data analytics projects never make it to production. Within financial services, given the levels of investments AI will require, and the vast amounts of data in daily scope, AI investments will be costly – especially if they don’t go on to deliver the efficiencies sought. However, finance is also fighting in an incredibly competitive space – and the businesses which can gear up to adopt AI at scale will be at a tremendous advantage. This tension between risk/ reward, combined with often multinational and complex regulatory landscapes and variances in data compliance demands, make widespread, institution-wide AI adoption high value but highly complicated. This puts yet more importance on good preparation. We have found that one in five businesses which start with a good data audit have gone on to deliver efficiencies – this significantly lowers the risk factor even for cautious and complex financial businesses.
It’s tempting to be swept up in the excitement around AI’s potential, especially at a time when finding business efficiencies is top of many boards’ minds. However, rush into projects without good planning and people may be simply rushing to failure. As we have seen when working with financial services data in numerous use cases, the sheer variety of financial data and its many uses within an organisation makes AI adoption complicated. However, with good planning and clear objectives on how AI introduction can optimise services, many financial organisations can start to deliver greater business benefits. An AI can only, and will only ever be as good as the source data it is fed, and the prompts from the business it is given. A strong data foundation will get businesses at least halfway to their AI wonderland.