[DSC DACH 24] Can AI and Automation Succeed on a Diet of Junk Data? - Sonja Vucic

Poor-quality data poses significant risks to the reliability and trustworthiness of automation and AI solutions. To mitigate these risks, organizations must prioritize building a strong data foundation. This requires a commitment to data governance and data quality from senior leadership and the establishment of a data-centric culture throughout the organization. By fostering collaboration between IT and business units, implementing robust data governance practices, and prioritizing data quality, organizations can create AI solutions that deliver value and are trustworthiness

1. Can AI and Automation Succeed on a Diet of Junk Data? Sonja Vučić NLB Banja Luka

2. Risks of Using Poor Quality Data Inacurrate predictions Inadequate business decisions Issues with processes Customer dissatisfaction Non compliance with regulations

3. Responsible AI

4. Strong data foundation for trustworthy AI

5. Illustration https://dataedo.com/cartoon M odel governance and testing play a big role in enabling correct results of AI , but quality of data is essential .

6. Strong data foundation for trustworthy AI Framework - Processes , Roles and Responsibilities Data Quality Dimensions and Data Quality Metrics Illustration https://dataedo.com/cartoon

7. Strong data foundation for trustworthy AI

8. Strong data foundation for trustworthy AI Illustration https://dataedo.com/cartoon

9. Strong data foundation for trustworthy AI Data quality rule # Rule Completeness 1 Must not be a null Accuracy 1 Email address is a real address and does not have an 'undeliverable' kickback Punctuality 1 E - mail address must be in entered as soon as the information is obtained by the customer Consistency 1 E - mail for Customer Number 12345 in System A must match Customer Number 12345 in System B Validity 1 2 3 4 5 6 Must have an '@' sign Must have only one '@' sign Must end with ".XX+" (2 or more characters after the "dot") Must not have spaces within the email address " " Must not have two dots "." next to each other Must not have more than one "@" sign

10. Strong data foundation for trustworthy AI Data Quality metrics – KPI Calculation example : A table with 1,000,000 records 5 0,000 of those records violate a Data Quality Rule Data Quality Score is 9 5 % (9 5 0,000 records “pass”) S et Thresholds (targets) for Data Quality scores , according to the needs for quality or according to criticality of the data

11. Strong data foundation for trustworthy AI Framework - Processes , Roles and Responsibilities Data Quality Dimensions and Data Quality Metrics Illustration https://dataedo.com/cartoon Cross-functional Collaboration Fostering Data- first Culture Accross the Organization

12. Thank you