Ensuring Data Quality in AI Projects: A Conversation with Stephanie Wiechers

In this episode, Erwin de Werd and Stephanie Wiechers explore the crucial role of data quality in AI and data projects. They discuss practical approaches to maintain high accuracy, the challenges of testing AI with AI, and the importance of human oversight to achieve reliable results.

Key Topics:

• The impact of messy data on AI output and decision-making
• Strategies for achieving 95% data accuracy for automation
• The process of data enhancement using AI and rule-based systems
• Testing AI models: AI-to-AI vs. human review approaches
• Cost and time considerations in data quality verification
• The ongoing progress: from 85% to over 95% accuracy
• The collaborative role of humans and AI in data validation
• Future outlook: the importance of human involvement for reliable AI

Timestamps: 00:00 - Introduction: How messy data costs industries billions 00:41 - Importance of data quality in AI and reporting 01:25 - Common issues with data errors impacting insight generation 02:17 - Automating error detection and correction in databases 02:58 - Client quality expectations and the 95% accuracy benchmark 03:26 - Achieving and validating 95% accuracy in AI models 04:01 - Using AI and internal rules for data enhancement 04:41 - Challenges of testing AI with AI and the need for human validation 05:56 - The risk of relying solely on AI for quality checks 06:37 - Human review as a reliable fallback 07:03 - The four-step process for data validation 08:25 - The iterative role of human review and AI learning 09:06 - Balancing internal and outsourced validation efforts 10:17 - Outsourcing testing versus internal validation challenges 11:13 - Current progress: surpassing 85% accuracy 12:00 - Upcoming guest episode and future projects

Resources & Links:

• PeerStop

Connect with Stephanie Wiechers:

• LinkedIn

Note: Stay tuned for our next episode featuring a special guest from the field discussing real-world data projects and best practices.