AI data-analysis prompts — turn AI into a junior analyst

Eight prompts that work when you paste a CSV, SQL result, or data table after the prompt.

## 1. Describe + sanity-check
```
Describe this dataset in 5 bullets: what each column is, the row count,
any obvious anomalies, what scale the numbers are in, and any
suspicious values worth flagging.
```

## 2. Top-N + summary stats
```
Show the top 10 rows by <column>. Then summary stats (count, mean,
median, min, max, p95) for that column. Then a one-line interpretation.
```

## 3. Cohort segmentation
```
Segment this data into 3-5 cohorts based on <column or behavior>.
Show each cohort's size + key descriptive stats + a one-line
characterization.
```

## 4. Anomaly hunt
```
Flag any row that's an outlier in any column (3+ stddev from mean,
or otherwise unusual). For each flagged row, explain why it's
suspicious.
```

## 5. Trend analysis (time series)
```
Identify the trend in <metric> over <time column>. Distinguish:
overall direction, seasonality, breakpoints. Give a one-paragraph
narrative an exec would understand.
```

## 6. Cohort-vs-cohort comparison
```
Compare cohort A (<criterion>) to cohort B (<criterion>) on these
metrics: <list>. Show a side-by-side table. End with the single
biggest difference and its likely explanation.
```

## 7. Hypothesis generation
```
What 3 hypotheses does this data suggest? For each, what additional
data would you need to confirm or reject it?
```

## 8. SQL drafting
```
I want to answer: <question in plain English>. Schema is: <paste
DDL>. Write the SQL. Explain the WHERE and GROUP BY choices.
```

## Tip
Always verify the AI's specific numerical claims against the source data. AI is a research assistant, not an oracle.