Isaac Song

Turning AI potential into real outcomes.

Director, People Systems & AI at WPP Media · Ex-BHP · I've deployed RAG, MCP, and LLMs inside global enterprises. This is where I break down how real companies are doing the same.

Leadership

AI Vendor Concentration Risk: What Leaders Need to Decide Before an Agentic Platform Runs Core Operations

Ryanair's five-year Google Cloud deal, built to survive an outage like Delta's 2024 one, shows why agentic AI needs a resilience plan, not just a roadmap.

Logistics

UPS AI Logistics: How Route Optimization and RFID Sensing Actually Cut Delivery Costs

UPS ran ORION and a new RFID sensing network at real scale for a decade, with verified savings, but the same year it cut 78,000 jobs.

Retail

Lowe's Mylow: How an OpenAI-Built Assistant Answers a Million Home Improvement Questions a Month

Lowe's built Mylow on OpenAI's models to close a real information gap for DIY shoppers, and gave associates a version too, with measurable, uneven results.

Manufacturing

Caterpillar Cat Digital: How Predictive Maintenance Cuts Downtime Across 1.5 Million Machines

How Caterpillar built a data platform across 1.5 million connected machines to flag failures before they happen, and what an enterprise would need to copy the approach.

Financial Services

Morgan Stanley AI Debrief and DevGen.AI: What 98% Advisor Adoption and 9 Million Lines of Legacy Code Reveal About Purpose-Built Enterprise AI

Morgan Stanley deployed two distinct purpose-built AI tools: Debrief for wealth advisor meetings and DevGen.AI for COBOL modernisation. Here is what the numbers actually mean, what was not disclosed, and what other enterprises can replicate.