Good list. One thing that helped with prompts 2 and 3: tell Copilot the column headers by name and the format, e.g. 'Sales is in USD, Date is dd/mm/yyyy'. A lot of wrong formulas come from dates stored as text, and Copilot won't always flag it. Before trusting any formula it writes, test it on three rows you can add up by hand, including one edge case like an empty cell or a region spelled differently. For the outliers prompt, ask it to list the row numbers and the rule it used, so you can check the logic instead of just accepting the highlight.
Thanks, Charlie. Naming the headers and the format in the prompt is exactly the kind of detail that saves the most time, and the dates-stored-as-text trap is one I see constantly in training sessions. Testing the formula on three rows you can add up by hand, including one edge case, is a habit I’ll recommend more explicitly. Asking Copilot to list the row numbers and the rule it used for the outliers is also a good way to check the logic instead of trusting the highlight. Have you found other prompts where this kind of check catches errors?
Charlie from xSquads
Head of GTM @ xSquads (xsquads.ai)
Good list. One thing that helped with prompts 2 and 3: tell Copilot the column headers by name and the format, e.g. 'Sales is in USD, Date is dd/mm/yyyy'. A lot of wrong formulas come from dates stored as text, and Copilot won't always flag it. Before trusting any formula it writes, test it on three rows you can add up by hand, including one edge case like an empty cell or a region spelled differently. For the outliers prompt, ask it to list the row numbers and the rule it used, so you can check the logic instead of just accepting the highlight.