Claude for Data Analysis
Published by O'Reilly Media, Inc.
Analyze and verify real data without writing code
What you’ll learn and how you can apply it
- Direct Claude in plain English to profile an unfamiliar Excel file and report what it contains, how much it covers, what is missing, and what looks wrong in it
- Organize the work as a project rather than a one-off chat, so the data, the steps, and the results stay together and can be picked up again later
- Answer factual, trend, and ranking questions about a dataset by iterating with Claude, without writing code
- Verify an aggregate Claude produced without reading its code
- Turn a checked result into a shareable dashboard that makes the finding obvious to someone who was not in the room
Course description
Most data work starts with a spreadsheet from a colleague, an export from some system, or an attachment with no explanation. An AI agent changes how that work gets done. You describe the task; Claude writes and runs its own analysis against your actual file, reads the result, and corrects its own errors. Daniil Maslyuk shows spreadsheet-proficient professionals who don't write code how to direct Claude through a complete exploratory analysis, and just as importantly, how to decide whether to trust what comes back.
This course teaches tool-agnostic, repeatable ways of working. The whole session runs on one dataset: 120 years of Olympic results and 200,000+ rows in a plain Excel file, downloaded along with a prompts document from the session resources at the start. In a sequence of hands-on exercises, you'll set the work up as a project, profile the file to learn what it contains and what is missing, answer factual questions and follow up on the surprises, chart a century-long trend as a shareable dashboard, compare groups across the data, and build an all-time medal table.
You'll practice five checks that need no code reading: asking what looks wrong in a file before counting anything, asking how a number was counted before looking at it, asking which column an answer was grouped by, re-counting the same table with a different key, and putting the same question to several independent agents and comparing their answers.
This live event is for you because...
- You’re a business analyst, BI user, or Ops or marketing professional who’s comfortable with spreadsheets but has never written code.
- You work with data files other people send you and you have to get answers out of them.
- You want to hand real data work to an AI agent rather than just chat with it, and to judge when its answers can be trusted.
Prerequisites
- A computer with a modern web browser, or the Claude desktop app on Windows or macOS
- A paid Claude subscription (plans start at US$20/month, billed in local currency where supported) that the hands-on work runs on
- Comfort with spreadsheet work (formulas, pivot tables, basic charts)
- No programming experience required, and no development environment to set up
Course preparation:
- Sign in to Claude in your browser, or install the Claude desktop app (free)
- Activate a paid Claude subscription (plans start at US$20/month)
- Nothing to download in advance: the course dataset, a single Excel file, and a prompts document are provided as downloadable resources during the session
Recommended follow-up:
- Read Hands-On Large Language Models (book)
- Re-run the workshop workflow on a file from your own work—everything built in the course stays on your laptop and can be reused.
Schedule
The time frames are only estimates and may vary according to how the class is progressing.
Section 1: Meet your agent (10 minutes)
- Presentation: Asking a question versus handing off a task: how an AI agent differs from a chat assistant
- Demo: The whole loop in five minutes, from "What is in this file?" to a finished dashboard
Section 2: Set up the project and meet the data (20 minutes)
- Presentation: Why a piece of analysis needs a home; why profiling comes before analysis
- Hands-on exercises: Download the course dataset and prompts document from the session resources; start the project, hand Claude the Excel file, and have it explain what is inside; ask what looks wrong in the file before analyzing it
- Q&A
Section 3: Questions that get answers (20 minutes)
- Hands-on exercises: Which medalists have the widest gap between their first medal and their last?; ask which column Claude used to follow one person through the file; chase the surprise—who won medals in more than one sport, and what that says about the early games; chart how many years each sport ran before women were let into it, and publish it as a shareable dashboard
- Presentation: The working loop—state the task, check the result, give specific feedback
Section 4: The ranking trap and how to catch it (35 minutes)
- Hands-on exercises: Build the all-time top-10 medal table and ask Claude how it counted a team medal before looking at the numbers; put the same question to several independent agents and compare what comes back; count the same table again with a different grouping key and watch the total move; compare the table against the published record books, then merge the historical country codes and watch places two to four change
- Presentation: Verifying methodology without reading code—the loop’s check step done properly
- Q&A
- Break
Section 5: Make the finding land (15 minutes)
- Hands-on exercises: Ask for three or four different views of the same finding and pick the one that carries it; refine the one you picked until a stranger gets the point in five seconds; the loop’s feedback step
Section 6: Wrap-up and Q&A (20 minutes)
- Hands-on exercise: A last question this dataset cannot answer
- Presentation: How to keep using this working process on your own files—the same three steps, what to run in parallel, what to always double-check; what will have changed by next quarter, and which of today’s habits will still hold
Your Instructor
Daniil Maslyuk
Daniil Maslyuk is a seasoned data and AI practitioner who is passionate about helping businesses get the most out of their data. He is a Claude Certified Associated and teaches analysts how to direct AI agents through real data work and check what comes back. For his community contributions, Daniil has been recognized as a Microsoft Most Valuable Professional (MVP).