Power BI with AI, Claude Code and Cowork

Course overview

In 2026 business intelligence has moved beyond building reports into an era where AI can do the work for you. Yet most Thai organisations find that Copilot in Power BI needs Fabric capacity of F2 or above, or Power BI Premium. The Pro and PPU licences most people hold cannot use Copilot, so many AI-for-Power BI courses end up as watching the instructor demo instead of doing the work yourself.

This course takes a path everyone can follow from the first minute: Claude Code, Anthropic's agentic AI assistant, connected to Microsoft's Power BI Modeling MCP Server through the MCP standard, running directly on the learner's machine alongside the free Power BI Desktop. Learners use AI to read the model structure, write and validate DAX, edit many measures in a single command, set up RLS, generate documentation automatically and have AI build and edit report pages through Power BI Project files (PBIP and PBIR). The course finishes with executive reports in Word and PowerPoint produced with Claude Cowork, together with an AI governance framework for using AI safely with company data. (2 days, 6 hours per day, 12 hours in total, Intermediate level.)

What you’ll gain

Understand where Power BI, Microsoft Fabric and AI agents are heading in 2026, including the licensing conditions of each AI feature
Design a semantic model that AI understands, following AI-ready modelling, without relying on Copilot
Understand agentic AI and MCP, and install Claude Code and connect it to the Power BI Modeling MCP Server yourself
Direct AI through modelling work end to end, from reading the model and writing and validating DAX to bulk operations, RLS and documentation
Understand Power BI Project (PBIP), TMDL and PBIR, and have AI create or edit report pages through these files
Use Claude Cowork to clean raw data and produce executive reports in Word, PowerPoint and Excel
Design agentic workflows with human review points, and understand governance, security and the risks of AI with company data

Who this course is for

Data analysts and BI developers who already use Power BI and want to upgrade their skills for the AI era
Managers, planners and analysts who build Power BI reports regularly and want to work faster with AI
IT and data teams who need a safe approach for using AI agents with company data
Organisations without Fabric capacity that want to start using AI for BI on a controllable budget

Prerequisites

Experience building reports in Power BI Desktop, including Get Data, creating visuals and publishing
An understanding of tables and relationships, and experience writing basic DAX measures such as SUM and CALCULATE
Basic Power Query transformations and good working knowledge of Excel
Basic use of Command Prompt or PowerShell; no programming is required

Curriculum

Course Details

This course is the 2026 reformatted edition of Power BI with AI, Claude Code and Cowork. It is a 2-day hands-on workshop, 6 hours per day (12 hours in total, 09:00-16:00), at Intermediate level. Across both days learners build one continuous project, Sales Intelligence 2026, a sales and customer behaviour analytics system for executives, on a mock dataset with no connection to real company data. The hallmark of the course is

that everyone can do the work for real without Fabric capacity, because it uses Claude Code connected to the Power BI Modeling MCP Server running on the learner's own machine with the free Power BI Desktop. Learners need their own or their organisation's paid Claude account. The course focuses on using AI agents for modelling and reporting work; in-depth AI visuals and Python are covered in the institute's Visualization in Power BI and AI course.

Day 1 An AI-Readable Semantic Model and Connecting Claude Code

Section 1: The 2026 BI Landscape and the Truth About Licensing

Where Power BI, Microsoft Fabric and AI agents are heading in 2026
The agentic BI concept: when AI no longer just answers questions but does the work
Which AI features need Fabric capacity and which do not, with a decision map for organisations
Why this course uses Claude Code and MCP as its core

Section 2: Lab: A Good Semantic Model Is the Foundation of AI

AI does not read your data, it reads your description of the data, with an 8-point model checklist
Build a star schema, relationships, a date table and basic measures
TMDL View and DAX query view for changing many model objects in a single script
Three-layer verification (syntax, semantic, business) and the createOrReplace trap

Section 3: AI-Ready Modelling and AI Visuals Available with Pro

Naming, descriptions, synonyms, perspectives and hiding fields that work with every AI tool
An overview of AI visuals available with Pro: Key Influencers, the decomposition tree, anomalies and narrative
Q&A retires in December 2026, and the transition path
Lab: bring the sample model up to the AI-ready checklist

Section 4: Lab: Installing Claude Code and Connecting It to Power BI

Install Claude Code on Windows and check it is ready
Install the Power BI Modeling MCP Server and register it with Claude Code
Connect to the semantic model open in Power BI Desktop
Have Claude read the model structure live and prove it reads rather than guesses

Section 5: Workshop Day 1: Sales Intelligence 2026

Build an AI-ready semantic model from the mock dataset
Lay out the main dashboard used across both days
Connect Claude Code to the model through the MCP server
Have Claude summarise the model structure into a document

Day 2 Claude Code, MCP and Cowork from Modelling to the Executive Report

Section 6: Getting to Know Claude Code, Claude Cowork and MCP

Claude Code, Claude Desktop, Claude Cowork and Claude for Excel, and choosing the right tool
CLAUDE.md, permission modes such as default, acceptEdits, plan and auto, and the settings file
The MCP architecture: host, client, server, tools, resources and prompts
Local and remote Power BI MCP servers, and security checks before installing any server

Section 7: Lab: DAX, Bulk Operations and RLS with AI

Have Claude write DAX measures from natural language and validate them with a real DAX query
Create and edit many measures, rename across the model and add descriptions to every object
Create Thai translations and use transactions so changes can be rolled back
Set up row-level security and perspectives with natural language
Check the AI's work with three-layer verification

Section 8: Lab: Preparing Data with AI

Always profile first: let AI survey the problems before touching the data
Clean missing values, duplicates and date formats, including Buddhist and Gregorian years mixed together
Combine CSV and Excel files from several sources ready for Power BI
Which jobs belong in Power Query and which suit AI

Section 9: Lab: Building Report Pages with AI through PBIP

Save as a Power BI Project (.pbip) with PBIR and TMDL
The structure of definition.pbir, pages, page.json and visual.json
Have Claude Code build report pages from natural language or a mockup image
Check the result in Power BI Desktop and track changes with Git

Section 10: Documentation, Cowork and Governance

Generate a data dictionary and model handbook automatically, with team skills and slash commands
Use Cowork to produce executive reports (Word), presentations and Excel files from reusable templates
Agentic workflow: what can be automated, what needs a human in the loop, and three review gates
Permissions, PDPA, MCP server rights, prompt injection, hallucination and checking the numbers

Section 11: Master Workshop: Sales Intelligence 2026

Have Claude Code create measures in bulk, such as YoY, running totals and RFM segmentation
Tidy up the whole model and generate documentation
Take a new dataset for AI to clean, update the model and build report pages through PBIP
Produce the executive summary and presentation with Cowork, then present the work

Schedule & training options

For individuals — public rounds

No public rounds are open right now. Join the waiting list and we will contact you first when the next round opens, or ask us on LINE. Or call 02-570-8449 or 088-807-9770

Join the waiting list Ask about rounds

For organisations — in-house / private

Tailor the content to your team’s tools and projects
Your dates, at your office or live online
Quotation with tax ID for procurement

Corporate training quote

Instructors

Samit KoyomFounder & CEO, IT Genius Engineering

Frequently asked questions

How do you design a Power BI semantic model that AI can understand?

Start with a star schema, with fact tables at the centre, dimension tables around them and a proper date table. Give tables, columns and measures clear business names, add a description to every object, add synonyms for terms people use differently, and hide technical columns such as join keys. AI relies on how your data is described far more than on the data itself.

How do you connect Claude Code to Power BI Desktop through MCP?

Install Claude Code on Windows, then install the Power BI Modeling MCP Server and register it with Claude Code. Open your report in Power BI Desktop so its semantic model is running, and ask Claude to connect to the open model and list its tables, relationships and measures. That confirms it can read the real structure before you ask for changes, and at first you should use a permission mode that makes Claude ask before it changes the model.

What is PBIP and how does it differ from a PBIX file?

A PBIX is a single binary file containing the report, model and data. A PBIP, or Power BI Project, saves the work as a folder of text files, with the model stored as TMDL and report pages as PBIR, which is becoming Power BI's default report format. That means you can read the files, compare changes in Git and edit them with external tools, including having AI such as Claude Code create or modify report pages through them.

Who is the Power BI with AI, Claude Code and Cowork course for, and what should I know beforehand?

It suits data analysts and BI developers already using Power BI, managers and analysts who produce reports regularly, and IT or data teams setting guidelines for AI use with company data. You should have built reports in Power BI Desktop, understand tables and relationships, write basic DAX such as SUM and CALCULATE, have used Power Query and Excel, and be able to run basic commands in Command Prompt or PowerShell. You need your own or your organisation's paid Claude account.

What will I be able to do after the Power BI with AI, Claude Code and Cowork course?

You will be able to design AI-ready semantic models, direct Claude Code through MCP to write and validate DAX, edit measures in bulk, set up RLS and document models, have AI prepare data and edit reports through PBIP, and use Cowork to produce executive reports in Word, PowerPoint and Excel, with review gates and governance for AI use. Participants who complete the training receive a Certificate of Completion with its own number, which can be verified online.

Can the Power BI with AI, Claude Code and Cowork course be run in-house for our team?

Yes. It can be run in-house for your team, with examples adapted to your own work, such as models and reports structured like your organisation's BI work. You can request a quotation from this page, or contact us on LINE @itgenius or by phone on 02-570-8449.