Transforming Supply Chains with AI at Cisco

Key Concepts

• Supply Chain Transformation: Evolving from core systems to intelligent operations using data and AI.
• Causal AI: Forecasting method using external macro factors to predict volatility.
• Gen AI Applications: Utilizing large language models (LLMs) for process automation, decision support, and autonomous orchestration.
• Data Foundation: Establishing a single source of truth with robust data security and governance.
• Digital Workflows: Standardizing processes and creating digital records for AI integration.
• Universal Fulfillment Orchestration: A digital platform for managing and optimizing order fulfillment.
• AI Agents: Autonomous systems that learn and take actions based on data analysis.
• RAG (Retrieval Augmented Generation): A technique for improving the accuracy and reliability of LLMs by grounding them in a specific knowledge base.

Cisco's Supply Chain Overview

• Cisco, founded in 1984, is a networking and IT company focused on secure AI infrastructure.
• The company recently acquired Splunk for $28 billion to enhance its full-stack observability capabilities.
• Cisco's supply chain manages hardware with over 3,000 employees and a 10x larger manufacturing partner ecosystem.
• The supply chain is structured according to the SCOR model: Plan, Source, Make, Deliver, and Service.
• Cisco emphasizes circularity, harvesting about half of the spares from returned products.
• The supply chain transformation team focuses on strategy, M&A integration, security, risk management, sustainability, and AI adoption.

Operating Models and Product Segmentation

• Cisco uses segmented operating models to manage different product classes and customer segments.
• Configured-to-Order (CTO): 40+ billion dollars of revenue, longer lead times (7-30 days).
• Build-to-Stock (BTS): Simpler, lower-end products like the Meraki line, shorter lead times.
• Cisco invests in its component subtier, including semiconductors and optics, to control cost, security, and quality.
• Cisco Services has its own supply chain, replenished by both CTO and BTS operations, as well as harvested returned parts.

Strategic Priorities

• Cisco conducts an annual strategic process, assessing the impact of 10 forces (geopolitics, economy, technology, etc.) on its priorities.
• Core strategies:
  • Growth: Scaling new products and technologies to new markets (e.g., webscale, AI data centers).
  • Adaptability: Improving planning, responding to market changes, and managing quality.
  • Innovation: Focusing on product, manufacturing, and process innovation, especially AI adoption.
• Sustainability, workforce strategies, and operational excellence are foundational to Cisco's operations.

Supply Chain Transformation Journey

• 15 years ago: Massive ERP upgrade to a single global instance and a standard product data management platform.
• Building Insights: Ingesting B2B data from manufacturing, logistics, and supplier partners.
• Digitizing Processes: Creating digital workflows for component disposition, corrective action, and inventory visualization.
• Intelligence Phase: Embedding machine learning for decision-making, requiring good data.
• Key takeaway: A strong data foundation is crucial for successful AI implementation.

AI Use Cases at Cisco

• Inventory of 150+ AI use cases across the organization.
• Teams rooted in statistics (planning, quality) have been the most progressive with AI.
• AI Maturity Levels:
  • Diagnostic: Robotic process automation.
  • Predictive: Algorithms providing predictions for planners or quality leaders.
  • Prescriptive: Algorithms prescribing and taking actions.
  • Cognitive: Gen AI algorithms that actively learn and work autonomously.

Forecasting with Causal AI

• Challenge: Traditional forecasting models struggle with volatility and recency bias.
• Solution: Causal AI from Causal Lens, using external macro factors (CPI, supply chain indexes, sentiment on inflation, interest rates) to predict bookings patterns.
• Results: Double-digit increase in forecast accuracy for the enterprise business (a $16 billion business).

Quality Improvement with Machine Learning

• Cisco's cloud-based test platform collects real-time data during manufacturing.
• Failure signatures are used for early failure sensing, root cause analysis, and corrective action assistance.
• Data is shared with services and repair teams for troubleshooting field failures.
• Results: Double-digit reduction in avoidable RMAs (returns) by identifying software issues, reducing warranty expenses.

Gen AI Applications

• Initial focus on LLMs for summarizing unstructured text.
• Process Control Documents: A Gen AI solution with a RAG-based training model for accessing and understanding process documentation.
• Universal Fulfillment Orchestration: Using Gen AI to analyze backlog data and predict orders at risk of missing their promise date.
• Evolving Gen AI: Moving towards Gen AI agents leveraging ML for augmented decisions and autonomous orchestration.

Gen AI Evolution and Complexity

• Complexity Levels:
  • Level 1: Summarizing documents, meeting transcripts (LLMs are good at this).
  • Level 2: Gathering information about suppliers from public domain (requires parsing vast amounts of data).
  • Level 3: Augmented decisions and autonomous orchestration (requires context and structured data).
• Key takeaway: Context is crucial for training models and making accurate predictions.
• Investing in semantic layers and leveraging business logic from digital platforms to train models.

Q&A Highlights

• Tariffs: AI is not directly applicable to tariffs; accurate master data is essential.
• Upskilling: Training programs for executives, data scientists, and the general population.
• RAG vs. Fine-tuning: Using open-source LLMs with vector databases for fine-tuning.
• AI and Traditional Processes: Integrating AI as an assistant to improve decision-making, not necessarily replacing the workforce.
• Centralization vs. Decentralization: Balancing distributed decision-making with standardization and best practice sharing.
• Regrets: Not having a clear data security strategy early on, leading to data silos.
• ROI and Learning: Balancing experimentation with investments in high-impact use cases tied to core metrics.

Synthesis/Conclusion

Cisco's supply chain transformation is a journey from foundational systems to intelligent operations, driven by data and AI. The company emphasizes a strong data foundation, strategic use of AI (including causal AI and Gen AI), and a balanced approach to centralization and decentralization. Key takeaways include the importance of context in AI applications, the need for continuous learning and experimentation, and the value of integrating AI as an assistant to improve decision-making. Cisco's experience provides valuable insights for organizations looking to leverage AI to optimize their supply chain operations.