Context unlocks meaning

A finance export from InsightCorp, a coffee distributor, just landed on your desk. The column names are internal codes, meaningless to the AI without a key. Finance finally sent the dictionary.

The export is already attached. First, ask a question that depends on a cryptic column (e.g., "Which item had the highest rev_q4_adj last quarter?"). The first answer will be a guess. Then re-ask the same question with the column dictionary included, and compare the two answers.

Column dictionary: it_id is the internal item ID; it_nm is the item display name; rev_q4_adj is gross revenue, adjusted for end-of-quarter discounts; unt_sld is units sold; mar_pct_n is net margin percentage after discounts.

Which best describes how the AI's answer changed once you gave it the dictionary?

This exercise is part of the course

AI for Data Analysts

Hands-on interactive exercise

Turn theory into action with one of our interactive exercises

This exercise is part of the course

AI for Data Analysts

IntermediateSkill Level

4.8+

Start Course for Free

Set up your AI-augmented analyst toolkit. Learn where AI fits across the five-stage analysis cycle, master the GCSE prompting framework for turning vague asks into actionable recommendations, and choose the right way to connect AI to your data: flat files, MCP, or a governed semantic layer.

Exercise 1: The AI-augmented analyst Exercise 2: Get to know The Daily Grind Exercise 3: Same prompt, different answer Exercise 4: Writing prompts that work Exercise 5: Pick a prompt apart with GCSE Exercise 6: Vague vs. structured prompt Exercise 7: Connecting to your data Exercise 8: Flat file, MCP, or semantic layer?Exercise 9: Context unlocks meaning

Current Exercise

Move from raw data to insights you can defend. Interrogate data quality for fuzzy duplicates, missing values, and impossible timestamps; enrich raw fields with AI as both doer and advisor; then find insights worth chasing across trends, distributions, differences, and outliers, and verify each one before it reaches a stakeholder.

Exercise 1: The interrogation phase: data quality Exercise 2: What can AI flag on its own?Exercise 3: Interrogate the coffee shop data Exercise 4: Interrogate the AI's deduplication Exercise 5: Enriching your data Exercise 6: Extract intents from support tickets Exercise 7: Doer or advisor?Exercise 8: Finding insights you can trust Exercise 9: Where would you start exploring?Exercise 10: Drill into a finding

Turn findings into dashboards and stories that land. Compress dashboard discovery and prototyping from weeks to an afternoon, tailor data stories to the audience and decision in front of you, and protect against polished-but-wrong output with the S.P.O.T. verification framework.

Exercise 1: Building dashboards with AI Exercise 2: From interviews to user stories Exercise 3: Prototype and iterate Exercise 4: Telling data stories with AI Exercise 5: Same data, two stories Exercise 6: Honest framing vs. misleading Exercise 7: Story for the CFO Exercise 8: Verifiability and trust Exercise 9: Which S.P.O.T. check catches this?Exercise 10: Does this number feel right?Exercise 11: Spot the polished-but-wrong dashboard

Run a complete AI-first analysis on Board and Beyond, a US retail chain. Audit data quality, identify the enrichments a category manager would actually use, surface and verify a headline finding, build a dashboard that backs an expansion decision, and deliver a one-paragraph story to leadership.

Exercise 1: Capstone Project: Board and Beyond Exercise 2: Get to know Board and Beyond Exercise 3: Which mess can't wait?Exercise 4: What would a category manager actually use?Exercise 5: Verify a headline finding Exercise 6: Where should we expand?Exercise 7: Tell the story to leadership Exercise 8: Congratulations!Exercise 9: Bonus: Data Analysis with Snowflake Cortex