The Best Data is Clean Data: An interview with Beju Shah, Bank of England
“The fundamental, before we get to the applications and applying AI at scale, is that you must get your data correct. This includes clear identification, clear data and standards of structured data. If you build up that good foundation there then you have that necessary quality of data to input into more intelligent processes. If the data going into the processes isn’t good it’s not going to be very intelligent”
On this episode of the SSON podcast, Beju Shah, Head of Data Collection & Publication, Digital Platforms at Bank of England explains why having no data is better than having bad data. Coming from a background in financial regulation and hands on experience in AI, Beju provides a detailed explanation of:
- Why data shouldn’t be considered ‘strategic’: “Do you really want to make decisions on poor quality data and entrust it to a machine that may not be as smart as you are?” asks Beju. Before applying AI of Machine Learning, Beju explains that you need a clean data, and while getting it isn’t an overnight process, it is vital to achieving success.
- Why technology should be seen as a way to improve FTE efficiency and should not be about losing them as resources but freeing up employees to do other projects that they previously didn’t have time to do. For Beju, technology should be used bring our human talents and pursue our best and most meaningful work.
- With supervision for banking data is 80% digitised with PRA based data, how Bank of England plans to focus on cleaning up content and definitions to make the data more efficient so that people can report against it.
- How Bank of England is putting a “common plumbing” in place for global organisations to work with and, as the technology evolves and they outgrow what they have, they will reap the rewards of new advantages with open and web based platforms that allowing anyone to integrate into their system however they wish to do so.
- How, on the road to more real time granular reporting, only time will provide an understanding of the realisms about the applications of AI and where they can use it and where they can’t use it yet.