Data

Data may or not be the oil of the twenty-first century, but it certainly is the fuel that powers AI. And yet, organizations have failed to devote the necessary attention and resources to build data as a strategic asset. And if there is a single reason why AI has not delivered the promised value, it is this myopic view of data. For the longest time, organizations have treated data as an afterthought: the recent excitement about data lakes and the investments in them haven’t yielded much either.
The situation with banks is no different. If there is one (more) lesson to take from the tech firms, it is that for data to truly generate economic value for the firm, it has to be treated as front and centre of the business and operating model. As banks, big and small, are finding out, this is easier said than done. The biggest constraint is legacy. The current data infrastructure and data models have evolved through functional silos, archaic systems and decades of accumulation of disparate systems through mergers and acquisitions. To add to that, organizational data fiefdoms have resulted in patchy (at best) progress in standing up a data ecosystem that can support AI to drive digital transformation.

This is why I believe that data must be one of the top organizational priorities for banks if they expect to survive and thrive in the fintech onslaught. Some of the key data priorities that banks need to embrace right now are:

1. Cloud data platform: The last decade saw organizations splurge millions on Hadoop systems, swept by the ‘big data’ narrative. That is unravelling rapidly, as the on-premise Hadoop infrastructure is unable to scale, both in performance and costs. It is time to move away to the cloud, and not just from the cost and scalability points of view. The cloud also offers enterprises the ability to innovate, especially with event logs and streaming data, which are the backbone driving real-time digital customer engagement.
2. Data pipelines: Any data infrastructure is only as good as the data. It may sound obvious, but the reality is that most banks have not done enough at all to move data from multiple legacy systems to a centralized source. Banks have (and continue) to burn millions of dollars on building enterprise data lakes with the promise of creating a curated, centralized data repository for all analytical needs. Clearly, this has not happened with any degree of success (not just in banks, but across industries) – largely because the time, effort and focus required are just not available in most organizations. Eventually, the result is less of a clean data lake and more of a muddy data swamp. In any case, it is important to invest in building and maintaining the flow of data into the cloud-based centralized repository for it to be even useful.
3. Data harmonization: This involves more than just transfer of data; it needs a complete integration and harmonization roadmap across disparate systems. In a typical bank, customer data is sourced from multiple systems – many organic (e.g. account transactions; loan origination; customer care) and many inorganic (systems acquired from bank mergers and acquisitions). Integration across these systems to create a complete data view is necessary. This is perhaps the hardest challenge to surmount, less from a technical perspective but more from an organizational process perspective.
4. Data quality: The GIGO (garbage in garbage out) principle has never been truer than in the digital world. Much like data harmonization, this is a tough problem, and yet a necessary one to solve. In a not-to-distant future, when the banks will need to drive personalized engagement with customers and automate operations internally, good-quality data will be key to ensuring that these processes drive the expected value.
5. Data for AI: AI feeds on data, and good AI is one that learns and keeps getting smarter over time. To make this happen, well-documented and reliable training and validation datasets need to be created. They must not just be comprehensive but also have guards against reinforcing algorithmic bias.
6. Data governance: None of the above is remotely possible without strong data governance that operates at the bank-level (not functional group level) and takes responsibility for:
   1. Breaking down the trust barriers to share data across groups
   2. Establishing a data ethics framework to protect against bias
   3. Advancing the cause of a centralized, shared data ecosystem with the right security and regulatory compliance

It is obvious but worth repeating that these technology investments in infrastructure modernization, enterprise AI and data are not to be treated as sequential. The real competitive advantage for banks is in the how: the ability to execute on all three fronts, even as they are continuously evolving with and learning from the market dynamics and other macroeconomic factors.

Nitin’s conclusion:

Digital transformation in retail banking is at an inflection point, and we can also expect this to accelerate significantly in the coming years, even as the industry is going through profound changes triggered by the economic uncertainty arising from the coronavirus pandemic. Success in this difficult environment will require clarity in strategic direction, the ability to make the right bets in technology and people, and, above all, speed and agility in execution. One thing, though, is increasingly certain: the banks of tomorrow will look a lot less like the branch-based networks we have been used to and ever more digital, with a limited physical footprint, offering customers a truly seamless multi-channel experience at every stage of their financial journey.