By skyforbes One of the most useful prompting technique I’ve found is asking/testing “around” the results.
Identifying dependencies and causal factors, and then checking those separately. Asking “what else must be true”.
Example: If I’m doing financial analysis and projecting revenue. I’d ask the model to identify what drives the revenue (like number of customers) and ask it to explain how the number of customers should change. Or what metrics will change because they depend on the increasing revenue.
How about for you?