The AI Journal
Notes on safe AI adoption, workflow, and the privacy constraints that shape both.
Draw It Before You Automate It: What an Omelette Taught Me About Mapping Business Processes
Before you automate a business process, draw it. A ten-minute omelette exercise from Abby Covert’s Stuck? Diagrams Help shows how process mapping surfaces the one decision where time, money, and mistakes hide — long before any AI enters the picture.
Why Most AI Projects Never Get Deployed: What I Learned from Eric Siegel’s Talk at Google
I watched Eric Siegel’s talk The AI Playbook at Talks at Google, and one idea stayed with me: the hardest part of AI isn’t building the model, it’s getting it deployed. Most machine learning projects fail because data scientists and business leaders each assume the other owns that step. Siegel’s fix is to treat AI as a consulting project, not a technology install: start from the business goal, define exactly what you’re predicting, and measure success in profit and savings rather than accuracy scores. His “prediction effect” shows that even a modest model can transform results when applied at scale. In this article, I share his key points and what they mean for anyone working with AI in business. Even shorter (one line, for a post or meta description): Why most AI projects never get deployed, and what Eric Siegel’s Talks at Google lecture taught me about turning models into real business results.
ChatGPT vs Local LLM for Sensitive Business Data: How to Decide
A practical framework for Canadian businesses deciding between ChatGPT and a local LLM for sensitive work, covering PIPEDA exposure, hardware reality, costs, and when each tool is the right answer.
More posts on safe AI workflows, the tools worth using, and PIPEDA-aware adoption coming soon.