Financial institutions are testing the capabilities of artificial intelligence (AI) and its application to their operational workflows. For many of their use cases, AI is a yet-to-be-proven technology for both their third-party and in-house applications.
In its application to data management, however, the stakes are higher: the success of AI in automating processes related to data quality will have an impact on more than the immediate workflows of the chief data officer’s team. If AI can be fully harnessed to improve the data that enters institutions’ systems, the entire enterprise will benefit.Being able to trust a technology application to carry out such extensive processing is crucial because it can do the work that would ordinarily have taken scores of employees many hours to complete.
“AI is making things more operationally efficient, it’s reducing the effort that you have to put into different components of data management,” Ellen Gentile, enterprise data quality team leader at Edward Jones tells Data Management Insight. “For example, I can just tell the AI tool to write a data quality rule and that’s really powerful, especially when compared with a longer and tedious process for documenting the requirements, along with coding, testing and then, finally, deploying it.”
Calum Conejo-Watt, data governance and quality lead at a large European private bank, agrees.
“These tools are a lot smarter, based on machine learning algorithms, and they’ll learn over time to support the data that is feed through and be able to help organisations respond to data issues much more quickly,” Conejo-Watt explains.
Hybrid Approach
Unsurprisingly, the main driving force behind many AI innovations is the need to reduce corporate costs. But that’s not the only incentive, as A-Team Group Data Management Insight’s 19th September webinar will discuss.
Entitled “Augmented data quality: Leveraging AI, machine learning and automation to build trust in your data”, the event will see a panel of experts, including Gentile and Conejo-Watt, cast a detailed eye over why and how institutions are seeking to optimise data through AI technology.
Talking Points
Having data that can be trusted is essential, and AI offers potential like no other technology to ensure that, says Conejo-Watt.
Without such assistance “data processing would just slow down”, he explains, adding that should that happen, organisations would need to commit more resources to keep up with the amount of data they take on.
Webinar panellists will also discuss best practices for implementing augmented data quality processes and will examine case studies of its real-world application.
They will also consider some of the challenges involved. Augmenting data quality with AI tools is no simple feat, offers Gentile. The New York-based executive explains that getting the best out any AI model requires that the data and instructions given to it are effective.
“You really can’t just say to AI, just go do this stuff,” especially with respect to third-party applications, she tells DMI. “What is not always apparent is the pre- or prep-work you have to do with the data and the tool itself. Telling it to write a rule sounds great, but there’s a back end to it, and what happens is you must create the data sets, or you must profile the data.
“It’s not straightforward and easy,” she adds.
This is where data-team expertise comes in, she argues. Data professionals can provide the necessary backup to validate the rules on which the models will run. In that way, rules building should be treated like any other piece of software development process.
“When it comes to highly consequential data, if you are facing any regulatory reviews or audit scrutiny and AI is ostensibly coding volumes of critical DQ rules and the rules are wrong, they’re going to scrutinise your process accordingly,” she says. “You really can’t forget the software development lifecycle in the process and human involvement.”
There are other risks to applying AI models to data quality tasks. Filling in data gaps by extrapolating from other data points is something that Conejo-Watt sees offering a potential solution to incomplete data sets. Gentile, however, warns that this could pose dangers without a strategy and careful planning.
“You’ve got to be very careful when you do that, because that’s actually transforming data,” she cautions. “You’d still have to make sure you’re capturing and fixing the issue at their root causes and not simply constantly filling in the gaps. You’d have to have a process and a standard to have a human intervene or look at those issues” to ensure the cause of the data paucity is addressed.
- Conejo-Watt and Gentile will be joined by Gil Cross, platform and AI product manager at Xceptor during the 19 September “Augmented data quality: Leveraging AI, machine learning and automation to build trust in your data” webinar. The event will begin at 10:00am ET / 3:00pm London / 4:00pm CET and you can register your attendance here.
Subscribe to our newsletter
