Bring the data, don't ask for it
The safest way to use AI for analysis is to give it the material and ask it to work with that: summarise these notes, compare these survey answers, find patterns in this table. Asking it to supply facts from memory is where invented numbers and sources come from (lesson 3).
Summaries that stick to the source
Summarize the interview notes below for a product manager.
- Five bullets maximum, most important first.
- After each bullet, put the note number(s) it came from in brackets.
- Only use what's in the notes. If something is unclear, say so.
- End with one line: what these notes can't tell us.
Notes:
[pasted notes, numbered]Citations back to the source make every claim checkable, and the last line stops the summary from sounding more certain than the evidence.
Analysing data
For small tables (a few hundred rows), you can paste CSV or a table directly:
- Describe the columns and what each row is.
- Ask a specific question: "Which month had the highest refund rate, and how does it compare with the average?"
- Ask for the working (chain-of-thought), so you can check the arithmetic.
- Ask what else to check: "What would you want to know before trusting this conclusion?"
For large or important datasets, use a spreadsheet or analysis tool for the numbers, and AI to explain and summarise the results.
Research synthesis
When combining several sources:
- Paste or attach the sources and label them [A], [B], [C].
- Ask for agreements, disagreements, and gaps across them.
- Ask it to separate what the sources say from its interpretation.
Specialists in Ace
In Ask Ace, @Oracle (The Oracle, Data) focuses on data and patterns, and @Researcher (Academic Researcher) on sources and rigour. Adding documents to your knowledge vault lets Ace search them when you ask.
Try it
Paste a set of notes, reviews, or a small table you work with into Ask Ace and use the summary prompt above. Check two of the bracketed citations against the source.