Smarter Humans, Smarter Machines was the core theme of our closed-door #RefinitivSocial100 UK roundtable held in September 2019, with 10 of the UK and Europe’s most influential thinkers and thought-leaders on social media in the world of FinTech.
- Banks have been saying for a very long time that the data they have is messy and needs to be cleaned. It doesn’t matter if it’s going into AI, or any other system.
- There is a disjoint between C-suite executives and data scientists. The C-suite have grabbed onto AI and think it will solve all their problems. The data scientists, on the other hand, understand that there’s a lot of work to go before you get to those high end benefits.
- To read more about AI and the data quality challenge, download the Refinitiv report: Smarter Humans. Smarter Machines.
Hosted by Refinitiv CEO David Craig and Ben Shepherd, Chief Strategy Officer, and moderated by Amanda West, SVP Innovation Refinitiv Labs, we held a near 2 hour-long discussion on the future of artificial intelligence and the challenge of poor data quality in financial institutions. The main question that everybody was seeking to answer was: Is poor data quality hindering the deployment of machine learning (ML) by financial services companies?
How serious is the challenge of poor data quality in deploying machine learning in financial services?
Liz Lumley (@lizlum), Director of content and Fintech ecosystem at VC Innovations, kicked off the discussion by explaining that financial organisations have faced a data quality challenge since long before AI emerged:
“Banks have been saying for a very long time that the data they have is messy and needs to be cleaned. It doesn’t matter if it’s going into AI, or any other system”.
While this was accepted around the table, FinTech entrepreneur, Xavier Gomez (@xbond49), stressed that AI is putting the problem of bad data into stark relief:
“When we imply that the data might be wrong, we are of course implying that the ultimate decision you make, the automated, autonomous decision made by the AI, having gone through machine learning, could be wrong”.
FinTech and digital payments advisor, Neira Jones (@neirajones) agreed that using bad data in ML somewhat defeats the object. She added that the problem is compounded by an increasing expectation that data should be accessible and usable immediately:
“Years ago, when we had huge data warehouses… you would have time to clean up your data. Companies had many people doing just that. Now, we’re in an era where we’re moving to instant. We want real time payments – and dirty data is even more of a problem”.
Amanda West, SVP of Innovation at Refinitiv pointed the way forward:
“You actually have to go back to basics and adapt the data into a fit state in order to be able to apply any of the sophisticated machine learning capabilities to your business”. She added, “We’re now seeing a lot of conversations in the market about how to prep your data to make it readily consumable”.
Agreeing with this approach, Neira Jones cited a different, but equally compelling, reason for cleaning up your data:
“Accuracy is also one of the fundamental principles of GDPR, so data needs to be clean. We’ve spent a lot of years not cleaning it. Well, now we simply have to do it.”
What is the cost of bad data?
Building on Niera’s point about GDPR, Xavier Gomez highlighted the legal risks that companies face by retaining or using bad quality data:
“You need to ask yourself, ‘Has the data been lawfully acquired? Do we need anonymize it?’ Complying with your legal requirements is an additional step in your data cleansing.”
Independent corporate governance advisor, David Doughty (@daviddoughty), echoed this view, suggesting that the cost of cleaning up data remains one of the primary blockers, regardless of legal obligations:
“The reason data remained dirty for so long is that the cost of cleaning it up just wasn’t justified. That position needs to change”.
Is the challenge overblown?
Timo Dreger of InsureTech Forum (@insurtechforum) had a different take on the question. While poor data quality is certainly an issue, he said, the scale of the problem is overblown due to massive levels of hype surrounding AI in financial services.
“Lots of start-ups say they are using AI. They say “we’re a tech company, we do things differently”, but when you actually look inside, they’re doing it manually.”
Timo quoted a 2019 survey from MMC Ventures which found that nearly 40% of European AI startups don’t actually use AI.
Smarter Humans. Smarter Machines. brochure from the #RefinitivSocial100 Breakfast Roundtable.
David Doughty agreed about the hype surrounding AI, but pointed a finger of blame at the board room:
“There is a disjoint between C-suite executives and data scientists. The C-suite have grabbed onto AI and think it will solve all their problems. The data scientists, on the other hand, understand that there’s a lot of work to go before you get to those high end benefits”.
David Craig, CEO of Refinitiv, wrapped up this particular part of the discussion succinctly, saying that he understood the excitement surrounding AI, but was in agreement that hype sometimes overtakes reality:
“Cloud computing is offering a degree of agility that we’ve never seen before, if you can only harness it. But sometimes the hype [around AI] is too great and the reality takes longer to arrive. This is a reality of modern business”.
To read more about AI and the data quality challenge, download the Refinitiv report: Smarter Humans. Smarter Machines.
Visit our website that ranks the full #RefinitivSocial100. We invite you to follow and join the conversation via #RefinitivSocial100 on social media.