AI struggle story
How to Improve Business Processes with AI: Design “Where to Use vs. Not Use” LLMs Using Deterministic vs. Non-Deterministic Work
The key to AI-driven process improvement isn’t simply “embedding AI into operations,” but deciding **where to use AI—and where not to**. Because LLMs are inherently **non-deterministic**, placing them in parts of a workflow that should be deterministic can lead to unstable quality, unexpected behavior, and added complexity. This article explains a practical approach: map your end-to-end business flow, separate deterministic vs. non-deterministic steps, and select improvement targets using **effort (time cost) × implementation difficulty**—with a concrete example and a simple difficulty matrix to find quick wins.
Read moreAI Is a Brilliant New Hire Who Knows Nothing—Here’s How I Give Instructions
AI outputs go wrong mainly due to missing context. This article explains how to provide task-critical information clearly, work within token limits, and treat prompting like onboarding a smart new hire—practical tips you can apply daily.
Read more