How Standard Bank solved AI, data governance challenges
The accuracy of data used by AI processes is critical. Using inaccurate or contaminated data would likely render AI output useless.
Businesses are trying to solve AI and data governance challenges in isolation. However, the whole data eco-system needs to be part of the data governance process.
This is the view of Ruan Vlok, head Data Exchange, Data Platforms & AI, at Standard Bank of South Africa, who will be presenting on 'Data governance for effective AI throughput’, at ITWeb Artificial Intelligence 2019, to be held on 20 August, at The Forum in Bryanston.
He says his organisation is currently focusing on solving this problem on an enterprise level to ensure it matures as an organisation and not as individual teams. “A lot of focus is given to data engineering practices to ensure that the flow of data into business intelligence, management information and artificial intelligence (AI/BI/MI) processes are sound and re-usable.”
In addition, he says Standard Bank has embedded data governance across the data provisioning flow to ensure it is something that happens naturally and not in a forced way.
“Another significant area of focus is data ownership,” adds Vlok. “Without formal ownership, we cannot place the required accountability with the right owners. We have adopted the concept of co-ownership. Data is not a single individual’s responsibility, and we have formally defined all personas involved in data ownership. Their roles and responsibilities have been defined and agreed across the Standard Bank Group”
According to him, once all these personas function in harmony, an organisation’s data strategy and governance becomes part of everyday processes.
Vlok says that defining data domains is also a critical component of effective data governance. “Not all data domains are equal. They are treated differently depending on sensitivity as well as security cclassification. Applying the right level of governance based on domain and security classification is crucial.”
Speaking of how data governance can boost AI benefits, Vlok says in building an effective data provisioning and governance pipeline, teams building AI components will spend less time in preparing and interpreting data, and more time on what really matters.
“Having the correct metadata explaining what the data means, will ensure that the correct input is used for AI processes. Accuracy of data is crucial, and using it from the wrong or contaminated source could render AI outputs useless.”
During his presentation, Vlok will share the bank’s approach to solving the data governance and data provisioning challenge. “Our enterprise data architecture was shaped over a couple of years and we are very effectively executing on our AI/BI/MI strategies. I will share the benefits we realised in dealing this on a group level and the importance of data co-ownership.”
Finally, Vlok will share his organisation’s approach to data curation and how this benefits consumers of data.