Build a better data-interpretation prompt
Start with what the numbers actually show
This fictional dataset contains enough information to describe a change and calculate rates—but not enough to prove what caused the change.
Fictional quarterly workforce snapshot
A workforce program wants to understand a shift in training activity between the same quarter last year and this year. Staff are considering whether the trend should change outreach or training strategy.
120 participants · 42 entered training · 27 entered healthcare training
100 participants · 50 entered training · 38 entered healthcare training
35% last year · 50% this year
64% last year · 76% this year
Referral mix, eligibility mix, provider capacity, employer demand, outreach changes, funding changes, participant preferences
“Explain why healthcare training is growing and what we should do about it.”
Add four things that keep description separate from explanation
The goal is not to make AI timid. It is to make the answer useful by labeling what the data supports, what can be calculated, and what still requires evidence.
For data work, ask AI to separate what the data shows, what can be calculated, what might explain it, and what evidence is still missing.
A correct number can still support an invented story
Imagine AI returns this interpretation. What would you do?
“Healthcare training is growing because local healthcare employers are facing severe labor shortages, and the workforce program’s outreach strategy is successfully steering more participants into high-demand careers. The board should invest more training funds in healthcare.”
Describe first. Explain second. Decide last.
Choose one low-risk chart, table, or trend you already understand. Ask AI to describe the change, show the calculation, label possible explanations as hypotheses, identify missing evidence, and suggest what you would need to check before changing strategy.
What changed for you?
There is no score and you do not need to check every box. Notice whether any of these feel more true now than they did before you started.
The goal is not to become an AI expert. It is to become a more confident, careful user of a tool that may be useful in parts of your work.
You have practiced meeting follow-up, employer outreach, policy summaries, board briefing notes, and data interpretation. Across all five, the skill is the same: give AI a real source, name what it must not invent, ask for a useful output, and verify the result before it leaves your hands.