A decision framework.
GOOD SIGNS
The output is checkable Errors are visible rather than silent
The task involves language: writing, summarising, transforming
You have the context to provide Volume makes it worth the setup A first draft is genuinely useful
BAD SIGNS
The output cannot be verified Errors would be costly and unnoticed Precise calculation is required Current information is essential and the tool cannot search The task requires context you cannot supply Accountability cannot be assigned
THE KEY QUESTION
If the output were subtly wrong, how would you know?
If the answer is that you would not, that is the problem.
THE SECOND QUESTION
What is the cost of being wrong?
Verify in proportion.
TASKS THAT ALMOST ALWAYS SUIT
Drafting something you will edit Summarising material you provide Explaining something unfamiliar Generating options to choose from Reformatting and structuring
TASKS THAT ALMOST NEVER DO
Final decisions affecting people Anything you will publish without reading Calculations you will not check
