Live Webinar | Closing the Data Quality Loop with AI Agents

This webinar introduces Soda's new product, Soda Cleanse, which is designed to help teams remediate data quality issues by closing the loop between identifying bad data and fixing it. Chapters: 0:00 Intro & housekeeping 1:55 Soda overview & the fixing bottleneck 2:40 How Soda identifies issues (observability + contracts) 4:15 Incidents & the diagnostics warehouse 6:20 Soda Cleanse: an AI agent that proposes fixes 7:20 The four AI fix strategies 10:25 Demo: reviewing & approving suggested fixes 16:00 LLM-based corrections: a worked example 20:30 Giving the AI agent extra context 25:55 Q&A Mathisse explains that while many platforms can identify data quality issues, the primary bottleneck for organizations is the actual process of fixing them. Soda addresses this with an end-to-end approach: - Identification: Soda identifies bad data using ML-powered observability for anomaly detection and data contracts that serve as the single source of truth for quality requirements. - Diagnostics: When rules fail, Soda stores the failing records securely within the customer's own infrastructure (a diagnostics warehouse) to maintain data security. - Incident Management: An incident is created for each issue, assigning it to the relevant data steward to ensure accountability. Soda Cleanse acts as a record-level fixing layer that uses AI agents to propose corrections for failed records. - AI Strategies: The AI agents employ four strategies to suggest fixes: constant values, native SQL (for lookups), Large Language Models (LLMs) for context-dependent corrections, and manual edits. - Human-in-the-Loop: Suggested fixes are presented to data stewards in their inbox. Stewards can review, edit, approve, or reject these suggestions before they are written back to the source application or warehouse. - Audit and Security: The platform includes full audit trails to track actions, as well as role-based access control to define who is permitted to write back corrections. Book a call to learn more: https://soda.io/demo?utm_source=youtube&utm_medium=organic&utm_campaign=&utm_content=book_demo