Field notes
Clear thinking. Better automation.
Practical answers to the decisions behind an AI programme: where to start, what people should control and how to tell whether the work is paying off.
Short answers first. Methods you can use. External sources linked.
Why most AI pilots fail, and the four questions that predict whether yours will
Read the field noteThe reading list
What are you working through?
Choose a subject or search for a question. Every note opens with a summary and a guide to its sections.
6 of 6 field notes
Why most AI pilots fail, and the four questions that predict whether yours will
The 95% figure is real but widely misread. What it actually measures — and the specific structural choices that separate the pilots that reach production from the ones that quietly stop.
The automation boundary: deciding what AI is allowed to decide
The single document that separates automation that gets switched on from automation that sits in a sandbox — and how to write one.
How to calculate the baseline cost of manual work
The formula a CFO will accept, the mistake that invalidates most business cases, and why observed times beat reported ones.
Process discovery without a process-mining licence
Celonis costs more per year than most automation programmes. Here is how to find what to automate without it — and where the tool-free approach genuinely falls short.
Twelve questions to ask an AI automation vendor before you sign
Written to be used against us as much as anyone else. If a vendor cannot answer these clearly, that is your answer.
AI calling and outbound rules: US, Australia, UAE and India
What changes when an outbound workflow uses an AI voice — and the controls to build before it calls anyone.
Put it into practice
Turn a useful idea into a first decision.
The scorecard helps you compare workflow opportunities. Bring its results to a working session to examine the process, people and systems behind them.