Tau
AI for data analysts · 11 lessons

AI for data analysts who check every number

Use AI to write formulas and SQL, tidy messy data, explore results and explain them clearly, checking every number. For analysts and anyone who works with spreadsheets or dashboards.

Who it’s for: Data analysts, BI developers, business analysts and spreadsheet power users.

AI can write a formula, a SQL query or a chunk of Python in seconds. That's a real help for analysts. But it can still produce wrong numbers that look perfectly right, and this course is built around catching them.

To write formulas and SQL, you'll give AI your schema, which is the layout of your tables, along with some examples. Then you'll test what it writes on edge cases. You'll clean a messy dataset with AI as a partner, logging every change. You'll run an exploratory analysis and recompute the key results another way. Then you'll turn the analysis into a story people act on, with chart titles and caveats that match the evidence, no more and no less.

Later lessons go further. You'll explain why a KPI moved this week, after first ruling out data pipeline and definition changes. You'll make an AI-written notebook rerun cleanly next month, and choose forecasting and statistical methods with AI's help. You'll also write the definitions needed by natural-language BI tools, which answer data questions typed in plain English. Then you'll test them with questions you already know the answers to.

By the end you’ll be able to…

  • Write and debug SQL, Python and spreadsheet formulas with AI
  • Run an exploratory analysis and verify every AI-generated number
  • Set up natural-language BI and test it with a golden question set
  • Protect personal data and document AI-assisted analysis

Short on time? Start with the essentials

These 4 lessons cover the most useful ideas in this course. Each one takes about a minute.

Essential 1

What AI can and can't do with your data

Explain how AI writes code to analyse data, why it can still produce wrong numbers, and where the analyst adds value.

Essential 2

Write formulas and SQL from plain English

Prompt AI with your schema and examples to generate correct formulas and SQL, then test them on edge cases.

Essential 3

Clean a messy dataset with AI as your pair

Use AI to profile data, find duplicates, fix types and standardise categories, logging every transformation.

Essential 4

Run an exploratory analysis and check every number

Run AI-assisted exploratory analysis and verify key results by recomputing them independently.

Every lesson in this course

This course has 11 video lessons of about a minute each, from the basics to expert habits. Each one ends with a question, three takeaways and its sources.

Foundations 2 lessons

  1. What AI can and can't do with your dataEssential

    Explain how AI writes code to analyse data, why it can still produce wrong numbers, and where the analyst adds value.

    “AI can write the SQL in seconds. It can also be confidently wrong. Know which.”

  2. Write formulas and SQL from plain EnglishEssential

    Prompt AI with your schema and examples to generate correct formulas and SQL, then test them on edge cases.

    “'Revenue by region, last 13 weeks, excluding returns' becomes working SQL.”

Building skills 3 lessons

  1. Clean a messy dataset with AI as your pairEssential

    Use AI to profile data, find duplicates, fix types and standardise categories, logging every transformation.

    “'NY', 'New York', 'new york ', 'N.Y.': one city, four problems.”

  2. Run an exploratory analysis and check every numberEssential

    Run AI-assisted exploratory analysis and verify key results by recomputing them independently.

    “AI made consultants faster, but 19 points less accurate on the wrong tasks.”

  3. Turn analysis into a story people act on

    Use AI to draft an executive summary, chart titles and caveats that match the evidence, no more and no less.

    “Your chart is right and nobody acted on it. Fix the headline, not the data.”

Advanced 5 lessons

  1. Explain why a KPI moved this week

    Use AI to decompose a metric change into volume, mix and rate effects by segment, after first ruling out pipeline, tracking or definition changes.

    “Revenue fell 8% overnight. Real drop or broken pipeline? Check that first.”

  2. Make an AI-written notebook rerun cleanly

    Review and refactor AI-generated analysis code into a documented notebook with pinned inputs and simple tests, so it reruns next month with the same results.

    “If you can't rerun it next month, it wasn't an analysis. It was a one-off.”

  3. Forecast and test hypotheses with AI help

    Use AI to choose and explain forecasting or statistical methods, check their assumptions and interpret results correctly.

    “AI picked a t-test. Your data isn't normal. Here's how to catch that.”

  4. Write the definitions natural-language BI needs

    Write the metric definitions, synonyms and verified example questions that make natural-language BI (e.g. Copilot in Power BI, Databricks Genie) answer consistently.

    “Ask 'what's our revenue?' and get three answers. Definitions fix that, not AI.”

  5. Analyse personal data with AI, lawfully

    Apply data minimisation, pseudonymisation, approved tools and documentation when using AI on personal data, under rules such as GDPR (EU and UK) or HIPAA (US health data), and check what your organisation requires.

    “Uploaded the customer table to a chatbot? You may have just shared personal data.”

Expert 1 lessons

  1. Test your AI analyst with a golden question set

    Build a set of real business questions with verified answers, run it against natural-language BI or an AI agent, and set the accuracy bar it must pass before wider rollout.

    “Before anyone trusts "ask the data", test it: 50 questions with known answers.”

Apps we recommend

These are the tools this course uses, and what each one is best for. We have no partnership with any of them.

Hex

Best for: Collaborative notebooks with AI agents

AI writes and fixes SQL and Python in shareable notebooks and data apps connected to your warehouse.

Freemium

Copilot in Power BI

Best for: Natural-language reports in Power BI

Builds report pages, DAX and summaries from plain-English questions, grounded in your semantic model.

Paid

Databricks AI/BI Genie

Best for: Ask-your-data spaces on the lakehouse

Lets business users ask questions in plain English over governed tables that analysts curate and tune.

Free trial
Hands-on project

Deliver one verified analysis your team can rerun

Take a real question from your backlog. With approved tools, and synthetic or de-identified data if personal data is involved, use AI to clean, analyse and chart it, recompute every headline number another way, and write a one-page summary for the stakeholder who asked.

What you’ll make: A rerunnable notebook or workbook, a transformation log and a one-page stakeholder summary with caveats.

Check your own work

  • Every headline number was recomputed by a second query or method and matches
  • The transformation log lists each cleaning step and the rows it affected
  • A colleague reran it from scratch and got the same results
  • The summary states the main caveats and what the data cannot show
  • No personal data left the approved environment

Try a ready-made prompt

Copy a better prompt for a real task, and see why it works better than the version most people type.

Further reading and viewing

We’ve chosen these articles, research papers, videos and guides from other trusted sources.

Research · Harvard Business School · 30 min

Navigating the Jagged Technological Frontier

Field experiment showing AI boosts quality on some tasks and hurts accuracy on others. Essential for analysts.

Guide · Microsoft Learn · 15 min

Overview of Copilot for Power BI

Official docs on what Copilot can do in Power BI and how to prepare semantic models for it.

Course · DeepLearning.AI · 20 hours

Data Analytics Professional Certificate

Full analytics programme with hands-on labs on when and how to use generative AI in analysis.

Questions about this course

Can ChatGPT write SQL queries?

Yes, and often well, if you give it your schema, meaning the tables and columns, plus an example of the result you want. It can still get joins, filters or date logic wrong. So test every query on edge cases, such as empty values, duplicates and boundary dates, before you trust it. The course teaches this habit for SQL, Python and spreadsheet formulas.

How do I check numbers from AI data analysis?

Recompute the headline numbers another way. If AI gives you a total from Python, check it with a pivot table or a simple query. Look at row counts before and after each step, and question any result that seems too neat. The course makes this a habit, and its final project asks you to verify every headline figure before you share it.

Can I put personal data into AI tools for analysis?

Only in tools your organisation has approved, and only the data you need. Use pseudonymisation, which swaps names and IDs for codes, or synthetic data where you can. Rules such as GDPR in the EU and UK, or HIPAA for US health data, may apply. The course covers these as examples, so check what your organisation requires and record how AI was used.

What's in Tau's course for data analysts?

There are 11 lessons, each about a minute long. Every lesson is a short video, then one question, three takeaways and the sources. They run from writing formulas with AI to testing natural-language BI tools such as Copilot in Power BI and Databricks Genie. Your first 5 lessons in Tau are free, and Tau Pro gives you every course and lesson.

Get early access to this course

Tau opens soon. Join the waitlist and we’ll email you as soon as you can start this course.