AI Tools for Data Analysts - Make Coders Jealous · Lesson 3 of 3

AI Tools for Data Analysts - Make Coders Jealous

Use AI assistants to accelerate analysis while validating code, privacy, sources, and conclusions.

Lesson

1. Introduction

AI assistants can draft a SQL query, explain an unfamiliar formula, suggest a chart, or help document an analysis. Their output is a proposal, not evidence. Analysts remain responsible for the data permissions, calculation correctness, and conclusions delivered to stakeholders.

2. Core concepts

A useful request states the task, allowed input, schema, expected output, and constraints. Review generated code line by line, run it on a small non-sensitive sample, and compare results to an independent calculation. Never provide secrets or restricted records to a tool unless the organization has explicitly approved that service and data use.

3. Topics covered in this module

Prompt context, schema descriptions, and iterative task refinement
AI-assisted SQL, Python, spreadsheet formulas, and documentation
Verification against sample rows, tests, known totals, and source definitions
Privacy, hallucinations, bias, licensing, and human accountability

4. Syntax and how to read it

# Review an AI-generated result against a known invariant
assert cleaned['order_id'].is_unique
assert cleaned['amount'].ge(0).all()

Assertions encode checks that should hold for this dataset. In production, report validation failures with actionable context instead of silently dropping records.

5. Practical workflow

Ask for a draft with explicit assumptions, inspect the method and dependencies, test it on known cases, and record what was changed. If results disagree with an authoritative source, investigate rather than choosing the more persuasive answer.

6. Common mistakes

Pasting personally identifiable, confidential, or credential data into an unapproved model.
Accepting fabricated functions, incorrect joins, or unsupported explanations because the answer sounds confident.

7. Best practices

Use synthetic or minimized examples when asking an assistant for help.
Keep a human reviewer responsible for any published number or business decision.

8. Practice task

An assistant writes a query that doubles revenue after a join. List two checks you would run before accepting the result.

9. One possible solution

Check key uniqueness and row counts on both tables, inspect the join cardinality, and compare the joined sum to a trusted pre-join total.

10. Summary

Use AI assistants to accelerate analysis while validating code, privacy, sources, and conclusions. Use synthetic or minimized examples when asking an assistant for help. Keep a human reviewer responsible for any published number or business decision.

11. What comes next

The next module, Advanced Excel for Data Analytics, builds on this foundation and applies it to a new part of the analytics workflow.

Visual guide

A reliable AI Tools for Data Analysts - Make Coders Jealous workflow

Step 01 Frame
Define the question, input scope, and intended output.
Step 02 Work
Apply the method and validate assumptions as you go.
Step 03 Check
Compare results with known cases and communicate limitations.

Code examples

Turn a vague request into a verifiable task

TEXT

Given columns order_date, region, and amount, draft SQL for monthly revenue by region. State date assumptions and show a check for duplicate orders.

Expected output

A reviewable query proposal plus explicit assumptions and a validation idea.

Verify a generated count

PYTHON

expected_ids = set(source['order_id'])
actual_ids = set(cleaned['order_id'])
assert actual_ids <= expected_ids