Overview
Claude is a language and reasoning model that can help with structured data work: cleaning and exploring datasets, writing and explaining SQL, planning dashboards and turning results into clear business narratives. This course shows analysts how to use it as a capable assistant while keeping every number checked and every decision with a person.
Day 1 covers analytical prompting, exploratory data analysis and SQL generation. Day 2 moves to dashboard planning and BI requirements, interpreting BI outputs, recurring reporting, working alongside spreadsheets and BI platforms, privacy and compliance, and scenario modelling for decisions.
Who it's for
Professionals who work with data and reports and want to speed up analysis without losing accuracy. It is especially suited to:
- Business analysts
- Data analysts
- BI and reporting specialists
- Finance and operations analysts
Prerequisites
- An understanding of basic data concepts
- Experience of spreadsheets or BI tools
- Familiarity with interpreting charts and reports
By the end of this course you will be able to
- Apply Claude to ad-hoc exploratory data analysis, including cleaning, summarising patterns and spotting anomalies
- Generate, explain and check SQL queries from plain-language business questions
- Plan dashboards, KPIs and data models, and write specifications for BI and reporting teams
- Interpret charts and BI outputs and turn them into business-ready narratives
- Build repeatable reporting prompts and work alongside spreadsheets and BI platforms with privacy and compliance in mind
- Evaluate options and model scenarios with structured decision frameworks while keeping human judgement in charge
Course outline
Times show the standard running order and can be adjusted when we tailor the course to your team. The trainer stays available during breaks to help anyone finish an exercise.
Welcome & orientation 10 min
How the two days run and the sample datasets you will work with.
Claude's capabilities for analytics 30 min
What Claude can do with data: reading uploaded CSV and Excel files, running calculations by writing and executing code, producing charts and tables, and explaining results. Where it can mislead, and why every figure gets checked.
Try it: Upload a small dataset, ask three questions about it, and check one answer by hand.
Prompting foundations for structured data 35 min
The GCES method (Goal, Context, Expectations, Source) for data tasks: describing the columns, the definitions and the output you want. Asking for assumptions to be stated and for the working to be shown.
Try it: Write a GCES prompt for an analysis of a sample sales dataset and compare it with a bare question.
From business question to analytical prompt 35 min
Turning a vague question such as 'why did margin fall?' into a precise, answerable one, with the metrics, period, comparison and breakdown specified.
Try it: Rewrite three business questions as analytical prompts and run one.
Cleaning and preparing datasets 45 min
Using plain-language instructions to find and fix duplicates, missing values, inconsistent formats and outliers, keeping a record of every change and working on a copy of the original.
Try it: Clean a messy dataset, list each change made, and spot-check five rows against the original.
Patterns, trends and distributions 40 min
Summarising patterns, trends and distributions, producing charts, and comparing groups. Asking Claude for alternative explanations before accepting one.
Try it: Summarise the trends in a dataset with two charts and write down two alternative explanations.
Insights and anomalies 35 min
Identifying insights and potential anomalies: what is unusual, whether it is an error or a real signal, and how to test it before reporting it.
Try it: Ask Claude to flag anomalies in a dataset, then investigate two of them.
Writing SQL through conversation 45 min
Describing the tables and the question and getting a query back. Stating your database type, supplying the schema, and testing every query on sample data, never first on a live system. Claude cannot run queries against your database unless it has been connected through an approved connector.
Try it: Give Claude a sample schema and write three queries of increasing difficulty, then run them on the sample data.
Explaining and improving queries 30 min
Having Claude explain a query line by line, find mistakes, and suggest optimisation approaches such as filtering earlier, simplifying joins and using indexes. Checking suggestions against your database's own execution plan.
Try it: Paste a slow or confusing query, get an explanation and suggestions, and decide which to apply.
Translating business questions into SQL 30 min
A repeatable routine: state the question, confirm the definitions, generate the query, run it on test data and reconcile the answer.
Try it: Turn two business questions into queries and reconcile the results with a spreadsheet figure.
Close 10 min
Recap of Day 1 and what we will build tomorrow.
Welcome back 10 min
Questions from Day 1 and today's plan.
Outlining dashboards and visual layouts 35 min
Working from the audience and their decisions to a dashboard outline: which questions it answers, which visuals suit them, and how to arrange the page.
Try it: Outline a dashboard for a chosen audience and ask Claude to critique the layout.
KPIs and data models 35 min
Defining KPIs clearly, with formula, source, grain and owner, and sketching the data model that supports them.
Try it: Define five KPIs for your dashboard and sketch the tables needed to calculate them.
Specifications for BI and reporting teams 30 min
Turning the dashboard outline and KPI definitions into a specification a BI team can build from: requirements, data sources, filters, refresh needs and acceptance checks.
Try it: Write a one-page specification for your dashboard.
Interpreting charts, metrics and dashboards 40 min
Uploading a screenshot or export of a chart or dashboard and getting a description, key movements and questions to ask. Claude can misread values from images, so confirm important numbers against the underlying data.
Try it: Ask Claude to describe a dashboard screenshot, then verify three numbers against the data.
Business-ready narratives and summaries 35 min
Turning results into a short narrative for managers: what happened, why it matters and what to do, with the caveats stated.
Try it: Write a half-page narrative for a leadership audience from your analysis.
Recurring reporting and document generation 40 min
Template-based prompting for reports you produce again and again, and generating the supporting documents for BI processes such as data dictionaries and release notes. Saving the routine as a Project or Skill.
Try it: Build a reusable reporting prompt and run it on two different datasets.
Working alongside spreadsheets and BI platforms 40 min
Using Claude with Excel through the Claude add-in, and with BI platforms by exporting data, sharing screenshots or, where your administrator has enabled a connector, querying directly. Structuring data inputs, with clear headers, definitions and tidy tables, so output quality is high.
Try it: Restructure a messy export into a clean input and compare Claude's answers before and after.
Privacy, security and compliance 25 min
What data may be shared with Claude and in which account, anonymising and aggregating before upload, access controls and review, and the risk of instructions hidden in data or documents.
Try it: Classify the fields in a sample dataset as share, anonymise or do not share.
Scenarios, options and human judgement 45 min
Using Claude as a co-pilot for decisions: scenario modelling and impact assessments, comparing options with structured decision frameworks, and keeping human judgement in charge of the final call.
Try it: Model three scenarios for a business decision, compare the options on a decision matrix and challenge Claude's recommendation.
Close: summary & next steps 10 min
Recap of the two days and what to try first at work.
What you'll need
Claude
- A Claude account (a paid plan is recommended). Analysing files and creating charts needs code execution and file creation to be switched on; on Team and Enterprise this is an administrator setting.
- For the BI session, the Claude add-in for Excel is installed from Microsoft AppSource, which your administrator may need to allow. Any BI or database connectors are optional and depend on your organisation.
Preparation
- A laptop with Excel or a spreadsheet application.
- Sample datasets and SQL scripts are provided so queries can be tested safely, without touching live systems.
Practice files are provided for every exercise. You are welcome to use your own everyday documents where your organisation's policy allows.