AI No Magic Wand for Augmented Data Quality: A-Team Group Webinar

Artificial intelligence (AI) is often seen as a magic wand for many data operations but when it comes to augmenting data quality processes, the technology’s potential comes with a few caveats.

Trusted data is critical to the smooth running of financial institutions’ operations. AI has shown incredible capabilities to ensure that, especially in such tasks as anomaly detection and filling gaps in datasets.

But without proper planning, organisations could create issues in implementation that are tougher than the ones they are seeking to solve, A-Team Group Data Management Insight’s latest webinar heard.

Preventing Bias

Calum Conejo-Watt, data governance and quality lead at Lombard Odier Asset Management, observed that at its most fundamental level, AI for ADQ faced a chicken and egg situation. AI needs good quality data to work optimally and without error. But that’s a stretch if it is being deployed to improve data in the first place.

One answer to this conundrum – and also to forefending against data bias in the  AI models – is to keep humans in the loop, Ellen Gentile, enterprise data quality team leader at Edward Jones, articulated in a sentiment that resonated with the rest of the panel.

For many companies, that means ensuring that core data management practices are up to scratch along with the formulation of strong governance and ethical policies around AI.

Poll Results

A poll of market participants, data consumers and vendors that tuned into the webinar illustrated the appetite for AI. A plurality of respondents said they were deploying the technology to cleanse and validate data. Those came ahead of application to use cases such as regulatory compliance. That finding was bolstered by another poll, which found data quality time to value was the main challenge for respondents.

AI’s appeal, however, is in its ability to tackle the challenges of ensuring good data quality. From making work easier across the enterprise and eliminating human touch points that can slow the data pipeline to improving data observability processes, AI-driven ADQ is adding value in multiple ways, said Gil Cross, platform and AI product manager at Xceptor.

Driving its implementation is the increasing complexity of data and growing use cases for it, which means that there needs to be more treatment of data when it comes into organisations. Because financial institutions operate in such a heavily regulated industry, data quality is paramount, Cross added.

Consequently, firms are taking advantage of faster ROIs that the technology now offers.

That’s also resulted in changes in the profile of users of AI. While the earliest iterations of these models were applied to front office functions where it was committed to revenue-generating activities, arbitrage and risk analysis, middle and back offices have since seen an “amazing explosion” in its adoption.

Methodical Planning

In concluding remarks, the panel agreed that while the challenges inherent in AI implementation in ADQ could be overcome through methodical planning, IT firms had also to overcome cultural obstacles, not least because such an endeavour would require investment.

Adoption of the technology could be difficult to get past teams that had little idea of what benefits would result, especially if change posed a risk to jobs and work patterns. An implementation roadmap also needed the support of management, who would be financing the project.

All of this argued for companies to take a holistic approach to their AI ADQ programmes by first understanding their data needs, the use cases to which it would be applied and the benefits they hoped it would bring. It would be prudent to conduct a proof-of-concept programme and to take time to educate staff about the reasons for embarking on such a path and the potential benefits of doing so.