Webinar: The Journey Towards Predictive FSQA Management

This webinar, titled "The Journey Towards Predictive FSQA Management," delves into how AI and digital technologies are revolutionizing food safety and quality assurance (FSQA). Featuring industry experts Geert Van Kempen from Veeva and Giannis Stoitsis from Agroknow, the session, facilitated by Nikos Manouselis, explores the acceleration of AI deployments in food risk prevention since 2021. The discussion highlights the transformative impact of AI, not just in modeling and technological enablement, but also in data aggregation, combination, and the development of user-friendly front-end applications like chatbots. The speakers emphasize that while AI is rapidly changing industries, the core challenge in FSQA is to demonstrate its practical value to an audience primarily focused on preventing food safety incidents, rather than technology for its own sake.

The conversation progresses by addressing common questions from the food industry, such as what peers are doing with AI, practical applications, and the challenges and pitfalls of adoption. Geert Van Kempen shares insights from Veeva's Product Summits, revealing that a "predictive state" in food safety is a high priority for industry leaders like Nestle, Mars, and Pepsi. These companies view it as a critical journey that cannot be undertaken alone, fostering collaboration within the non-compete space of food safety. Giannis Stoitsis corroborates this, stressing the need for robust data and software infrastructure to enable risk prevention, acknowledging the complexity while also pointing to valuable basic steps that can be taken. A significant point of agreement is that while the will to adopt predictive FSQA is strong, the practical implementation faces substantial hurdles, particularly around data integration and standardization.

The webinar then dissects three practical use cases for AI in FSQA: mitigating external risks, monitoring internal production facilities, and managing supplier-related risks. For external risks, AI helps identify unexpected new threats from global supply chains by processing public data like recalls, border rejections, and inspection results to provide actionable alerts and forecast incident trends. Internally, AI aids in developing methodologies to assess factory-specific risks by integrating diverse data points—from audit findings and non-conformances to environmental monitoring and even non-classical quality data like absenteeism and training records—to generate leading indicators. Finally, for supplier risk management, AI combines external risk intelligence with internal data (audit performance, lab results) to create dynamic risk profiles, forecast potential hazards, and prioritize preventive measures, thereby shifting from reactive to proactive supplier management. The discussion concludes by reflecting on the broader adoption curve of AI, the level of trust required for automated decision-making (e.g., automatically delisting a supplier), and the call for pragmatism in starting the AI journey, focusing on quick returns while building foundational capabilities.

Key Takeaways:

• Accelerated AI Adoption: AI and predictive analytics have seen significant acceleration in food risk prevention since 2021, with applications like chatbots transforming how industries interact with vast amounts of knowledge.
• Three Pillars of Transformation: AI's impact spans three core areas: advanced modeling, aggregation and utilization of diverse data, and innovative front-end applications that facilitate human interaction with these technologies.
• Industry Prioritization: A "predictive state" in food safety is a top priority for leading food manufacturers, who see it as a significant enabler for improving food safety management from a reactive to a proactive stance.
• Collaboration is Key: Food safety is widely recognized as a non-compete area, fostering collaboration among companies to share data and insights, as no single entity can tackle the problem alone.
• Data Infrastructure Challenges: A major hurdle is the collection, integration, and harmonization of data, both external (public sector information from diverse sources, formats, languages) and internal (disparate LIMS, ERPs, quality systems). Data silos and lack of standardization (e.g., for food, hazard, or region classification) hinder effective analytics.
• Focus on Leading Indicators: While traditional quality data provides lagging indicators, the true value of predictive FSQA lies in identifying leading indicators. This requires integrating non-classical data points like absenteeism, changeovers, and training records to anticipate risks.
• Practical Use Case - External Risk Mitigation: AI can provide actionable alerts and forecast incident trends by analyzing global public data (recalls, border rejections, inspections), helping food safety teams design risk-based testing programs and share knowledge across large organizations.
• Practical Use Case - Internal Facility Monitoring: AI helps assess risks within manufacturing environments by combining audit findings, non-conformances, analytical results, and environmental monitoring data to identify potential issues before they escalate.
• Practical Use Case - Supplier Risk Management: AI combines external risk intelligence (forecasted hazards, non-compliances) with internal supplier performance data (audits, lab results) to create dynamic risk profiles, enabling proactive communication and targeted preventive measures.
• Digital Pipeline for Agility: The industry aims to create a "digital pipeline" where risk information flows seamlessly from external intelligence platforms into internal quality management systems, enabling agile responses to emerging hazards and informing updates to HACCP plans.
• Trust in Automation: A critical open question is the extent to which organizations will trust AI algorithms to make fully automated decisions with significant economic and public health impacts, such as automatically delisting a supplier.
• Start Pragmatically: Companies are advised to start their AI journey with pragmatic initiatives that yield a quick return on investment, focusing on learning and building capabilities over time, as the benefits are significant and not adopting will lead to falling behind.
• Integrate Solutions, Not Just Data: Beyond integrating data, there's a growing realization that integrating software solutions themselves is crucial to effectively support food safety and quality assurance experts, moving beyond "software silos."
• Preventable Crises: Examples like lead contamination in cinnamon or allergen contamination in spices highlight how predictive technologies, by identifying increasing incident frequencies or known hazards, could have enabled proactive measures and prevented costly recalls.

Tools/Resources Mentioned:

• Veeva Systems: Described as a provider of cloud-based Quality Systems, specifically for the food and beverage industry in this context.
• Agroknow: The channel host and a company specializing in AI and digital technologies for food safety and quality assurance.
• Predict System (USFDA): A pilot program by the USFDA (started 2019) using data and machine learning algorithms to highlight risky seafood imports, now heavily deployed in US importing ports.
• Data Lake: Mentioned as an infrastructure investment for consolidating data, but highlighted as insufficient without data standardization.

Key Concepts:

• Predictive FSQA Management: Shifting from reactive food safety and quality assurance (based on past incidents) to a proactive approach that anticipates future risks using data and AI.
• Leading vs. Lagging Indicators: Leading indicators predict future events (e.g., absenteeism as a sign of stress leading to mistakes), while lagging indicators describe past events (e.g., audit findings, non-conformances). Predictive FSQA aims to leverage more leading indicators.
• Non-Compete Space: The concept that certain areas, like food safety, are universally beneficial and should encourage collaboration among competitors rather than competition.
• Data Silos: Disconnected data storage and management systems within an organization, hindering comprehensive analysis and integration.
• Data Harmonization/Standardization: The process of making data from different sources consistent and compatible, crucial for effective AI and analytics.