Final exam: AI for data analysts

Read in

1. An analyst pastes a spreadsheet of patient visit records into a public chatbot to build a summary. What is the main problem?
The summary might come out too short to be useful to the clinic's leadership teamIdentifying health information was shared with a tool not approved for itPublic chatbots cannot read tables, only plain paragraphs of typed textThe spreadsheet has too many columns for the chatbot to summarize accurately
2. A query drafted by AI returns a total that is much larger than the finance team's figure. What is the most likely first thing to check?
Whether a join is repeating rowsWhether the chart colors match the finance team's slide templateWhether the column names are too long for the report layoutWhether the AI tool was a free version or a paid version
3. The assistant's summary states a growth figure that you cannot find in your data. What should you do?
Remove it unless you can trace it to the sourceAdd a footnote saying that AI wrote the figure so readers can judge itRound it to a safer figure so the number is less likely to be wrongKeep it, because it sounds plausible and matches the trend you expected
4. Which prompt will most likely give a useful query?
Ask for a finished dashboard in one step and skip the query so you do not have to read SQLSay what you want in one line, like 'write SQL for my sales', and fix whatever comes backPaste the full customer tables with real values and ask the tool to find anything interestingGive the database type, table and column names, the question, and ask for an explanation of each join
5. Which task is most likely to stay human?
Labeling the chart axes and legends so the figures are easy to read at a glanceWriting a first-draft formula to calculate a rolling monthly averageReformatting a table so the columns line up with the report templateDeciding with a director which question the business should answer first
6. A bar chart you built starts its vertical axis at a high number. What is the concern?
The axis will be mistaken for a line chart and confuse the legendTall bars leave no room for labels, so the figures cannot be readThe chart will print poorly and use far more ink than necessarySmall differences may look much larger than they are
7. Your data came from an online survey. What question helps you spot bias?
Which chart template did the survey team use?How many responses did the survey collect in total?Who could not have answered it?How recently was the survey spreadsheet last saved?
8. A coworker asks how you used AI on a report. What is the best reply?
I would rather not say, since people might think the analysis is less my own workI had it draft the query, then I verified totals against the source and documented my assumptionsIt did all the work, and I just forwarded the output after a glance at the chartsI do not remember which parts it wrote, but the numbers looked fine to me at the time

Name to print on your certificate

We don't check identity, and we don't store your name: it lives only inside your certificate link.

Answer all 8 questions and enter a name to submit.