AI for Data Analysts masterclass

What an assistant knows and guesses about your data, how to describe data instead of exporting it, the quiet ways AI-assisted analysis goes wrong, and the rules to know before the first prompt.

1. What an assistant knows, and what it guesses · 12 minThe learner can separate what an assistant can reliably do for an analysis from what it can only guess about their data.
2. Describe your data instead of exporting it · 13 minThe learner can give an assistant a schema description and invented examples that are enough for the task without sending personal or confidential data.
3. Where AI-assisted analysis goes wrong · 13 minThe learner can name the common failure patterns of AI-assisted analysis and say which check catches each one.
4. Your data rules before the first prompt · 11 minThe learner can find and apply their organisation's rules on approved tools, data classes and retention before using AI on data.

The analyst's weekly work done with an assistant and checked: SQL from a specification and its review, spreadsheets, Python, cleaning, joins, exploration and charts that do not mislead.

1. Write SQL with an assistant from a specification · 14 minThe learner can write a query specification that leaves an assistant nothing important to guess.
2. Review SQL an assistant wrote · 15 minThe learner can review an AI-written query line by line against its specification and test it on data they can check by hand.
3. Spreadsheet analysis with an assistant · 13 minThe learner can use an assistant to build spreadsheet formulas and pivots and verify them on cells they can check.
4. Python and notebook analysis with an assistant · 14 minThe learner can get dataframe code from an assistant and add the checks that show it did what was intended.
5. Clean data with an assistant, and keep a record · 14 minThe learner can plan and apply cleaning rules with an assistant while keeping a log of what changed and why.
6. Join data without losing or multiplying rows · 13 minThe learner can plan a join with an assistant and prove it neither dropped nor duplicated rows.
7. Exploratory analysis with an assistant · 13 minThe learner can use an assistant to widen an exploratory analysis without letting it decide what the findings are.
8. Charts that do not mislead · 13 minThe learner can check an AI-drafted chart for axis, scale, selection and encoding choices that distort the data.

The types of AI tool analysts use, how to limit a tool that can reach your data, and how to choose and justify a tool for a brief, with costs looked up live.

1. Tool types for analysts · 12 minThe learner can describe the main types of AI tool for analysis, what each is for and where each tends to fail.
2. Tools with access to your data · 13 minThe learner can set limits on an AI tool that can read or query their data, so that a mistake or a manipulated instruction cannot cause harm.
3. Choose and justify a tool for a brief · 12 minThe learner can pick a tool type for an analysis brief and justify the choice on data rules, task fit, failure risk and cost.

Independent reconciliation, the statistics traps AI walks into, findings written as evidence, inference and gap, text and survey analysis, and chained steps with checkpoints.

1. Reconcile numbers independently · 14 minThe learner can confirm a result by a route that does not share the assumptions of the original analysis.
2. Correlation, cause and the stories AI tells · 13 minThe learner can spot when an assistant turns a correlation into a causal claim and rewrite it with the alternatives named.
3. Small samples, survivorship and other traps · 13 minThe learner can recognise small-sample, survivorship and denominator errors in an AI-assisted analysis and correct the claim.
4. Write findings with evidence, inference and gap · 14 minThe learner can write a findings section in which every claim is labelled as evidence, inference or a known gap, with its source.
5. Text and survey analysis with AI · 15 minThe learner can use AI to code open-text responses into themes and measure how far the coding can be trusted.
6. Chain analysis steps with checkpoints · 13 minThe learner can break a multi-step analysis into stages with a check between each, rather than asking for everything at once.

Building it into the team: reproducible notebooks, a shared library of definitions and prompts, a review process, recurring dashboards and an automated weekly report with a person checking.

1. Reproducibility and notebooks · 13 minThe learner can make an AI-assisted analysis reproducible, so that someone else can rerun it and get the same result.
2. A shared library of definitions and prompts · 13 minThe learner can set up a team library of metric definitions, context blocks and prompt templates that keeps AI-assisted work consistent.
3. A team review process for AI-assisted analysis · 13 minThe learner can design a proportionate review process that decides which AI-assisted analyses need a second person and what they check.
4. Dashboards and recurring reports with AI · 13 minThe learner can use AI to build and maintain dashboards and recurring reports without letting definitions drift or errors repeat each cycle.
5. An automated weekly report with a person checking · 16 minThe learner can design an automated weekly report in which AI drafts, automated checks gate, and a named person approves before it is sent.

A lab of AI-drafted charts with correct numbers and planted design problems. Practise finding them, then pass the certified round.

1. Lab briefing: how charts mislead · 12 minThe learner can name the planted problems the Spot the misleading chart lab uses and the question that exposes each.
2. Lab practice: spot the misleading chart · 15 minThe learner can complete practice rounds, identifying every planted problem in a chart and proposing a fair redesign.
3. Lab certified round · 11 minThe learner can pass the certified round of Spot the misleading chart under the same rules as the other Academy labs.

A checked analysis of your own data or the provided Larch & Lane dataset, reviewed by a person against an eight-item rubric, each item met or not met, with a verified-and-not list.

1. Capstone brief and rubric · 16 minThe learner can plan a checked analysis of their own data, or the provided dataset, against the published capstone rubric.
2. Capstone: the analysis and the verified-and-not list · 20 minThe learner can complete the capstone analysis and submit it with a verified-and-not list that a reviewer can check.

How the exam works and how the AI for Data Analysts certificate is earned.

1. The exam and the certificate · 10 minThe learner can explain how the AI for Data Analysts exam works, what it covers and how the certificate is earned.