Many of you reading this will either have first-hand experience with the challenges of achieving data-driven success within financial firms or will have a reasonable concern that such success will not come easily to your organization. It is certainly true that finding actionable insights within your organizations — and then actually taking meaningful action based on those insights — is not a trivial matter. And achieving this not once, as part of some special project, but consistently and as a matter of regular business, is an even more substantial goal.
These challenges are widespread, as is the difficulty in overcoming them. But it is not the case that these challenges are insurmountable: quite the opposite in fact. These are tractable problems, and the right combination of strategy and toolset can bring about great success: unlocking real business value within your organization's depth of knowledge and data. Indeed, your firm is full of valuable knowledge, valuable ideas, and valuable data. The problems arise connecting these together consistently, efficiently, and effectively. Those firms that achieve a consistent, scalable, efficient data science solution will be able to access previously inaccessible process efficiencies, and develop compelling new products and value propositions.
.jpg?width=600&name=matthew-henry-VviFtDJakYk-unsplash%20(1).jpg)
The challenges that face firms who seek data-driven solutions have not shifted dramatically in the past year. Though AI regulation at national levels has become a real rather than theoretical concern, firms grappling with that as an immediate, everyday problem tied to revenue-generating AI models are operating at the highest levels of AI maturity: Most of the financial services industry is far from such challenging heights.
What has likely come into focus for most is that AI solutions are neither fanciful vaporware nor revolutionary panaceas: They are instead focused, delimited elements within wider operations and corporate cultures. This reality was either learned directly, perhaps painfully, by individual organizations and leaders in the vanguard, or observed ruefully by those who have now started to engage with AI substantively. This perspective shows the maturing of AI as a tool within the financial services industry, and the recognition that, to leverage it effectively, will require substantial and ongoing investment in not only tools, but strategies and cultures. At this inflection point towards growing maturity, we see the ongoing importance of existing trends:
AI Pipelines Fully Operational
Initial efforts to deploy AI solutions faced many obstacles, but one of the most frustrating and commonplace was the difficulty faced by firms in taking a solution with demonstrated value as a proof-of-concept into the real world, especially if that involved directly connecting customers with the AI process. This blockage in ‘productionization’ impacted the credibility of AI within the industry, as money put in resulted in little revenue-relevant output and large backlogs of unproven designs. This challenge is now well understood to be common and widespread, and something that can be actively and effectively addressed as part of any AI strategy.
By combining the right tools with effective model review and implementation processes, ever more AI solutions will make it far enough to prove real world value to the business or fail — in either case becoming a normal part of business, rather than as vague potential that never makes it far enough to be evaluated. This will have the additional effect of helping business leaders delineate areas where AI solutions are most likely to be effective, based on real world examples in their own and peer firms, and areas where AI is not (yet) suitable for their own and their customers’ needs.
Conversion of End-User Computing (EUCs) Into AI-Adjacent Computing
The best built effective AI pipelines include substantial data cleaning and processing abilities in powerful and auditable environments. These can — depending on their design — be leveraged directly by non-AI teams. Firms which encourage and enable this effort can reap substantial benefits, without needing to deploy a single new model into production. Giving data-savvy or technical teams who have no immediate need for AI the right tools, platforms, and training will allow for the slow but definitive whittling away at the plethora of EUCs littered across many financial services firms. This approach will allow EUCs to shift from desktops into governed and enterprise-grade platforms without loss of agility. As a result we will see that, in the short term, many processes currently handled via EUCs will not be replaced by AI, but will instead simply be rebuilt in robust, transparent, and governable forms.
The Business Team AI Vanguard
Just as some teams will benefit from the upstream work of data cleansing, organizing, and ease of data access without needing to leverage AI, small pockets of business-embedded teams will be able to incorporate simple AutoML and off-the-shelf AI solutions to enhance existing work. In one part of the firm a dedicated AI team might create a real-time Next Best Product (NBP) machine learning solution that sends customers targeted offers within seconds of a transaction completing.
At the same time, in another part of the firm, business teams with some analytical confidence might incorporate a simple NBP plugin, hooked up to their internal data warehouse on a shared data science platform, to quickly generate indicative product areas for marketing teams to focus on across each individual country within a regional business portfolio. The resulting analysis could show up on a single slide of a fifteen page deck presented to regional product leads, sparking interest in their approach and conclusions, and ultimately influence the final decision made. The underlying analysis might be considered simplistic by a data science team, but it will be considered great value-for-effort to the business team who created it.