Helping colleagues get started.
WHAT TO TEACH FIRST
That it generates plausible text, not verified facts That context must be provided That the first output is a draft That everything factual needs checking
THOSE FOUR PREVENT MOST PROBLEMS
WHAT TO DEMONSTRATE
A vague prompt and its output The same request with context and specifics The difference is usually striking and it teaches more than explanation
WHAT TO WARN ABOUT
Confidential material Publishing unchecked output Accepting agreement as validation
WHAT TO PROVIDE
A short written policy: what may be used, what may not be sent, who to ask
A few prompts that work for your common tasks
WHAT NOT TO DO
Mandate use without training Ban it entirely, which drives it underground onto personal accounts with no oversight Assume people will work it out
THE PRACTICAL APPROACH
One session showing real examples on real work, plus a one-page policy.
That is more effective than any amount of general guidance.
