This repository contains the instructions and code for the 90 minute workshop.
This workshop uses an AI Scenario to help set the stage with a business domain and desired functionality: Using AI Agents to create a contract review agent.
NOTE: Since a lot of the material on that site is geared toward using Copilot, I need to mention I have only used the requirements as inspiration in this workshop, we do not use Copilot in this solution
The exercises in this workshop are to provide a hands-on experience to help you understand how you can start with content in unstructured data, process it and use agents to provide a system that can unlock the value for your users.
- Contract ingestion - learn to process a pdf using Azure AI Document Intillegence, identify and create metadata to use in filtering, modify embedded text to be more effective in retrieval.
- Risk identification and Clause suggestions - analyze contract clauses, comparing to a template contract and set of desired terms
- Contract comparison - compare two contracts
- Rewrite a contract with suggestions - create a new word document of a rewritten contract that starts with an uploaded contract and takes into account a template contract and the desired terms
- GET SOME HANDS ON EXPERIENCE WITH SEMANTIC KERNEL AGENTS
Please install the software ahead of the workshop:
- VS Code
- Python 3.12 and PIP, also recommend installing the VS Code Python Extension - this will allow you to debug and step through code later
- (Optional) Prompty Extension - this will allow you to test prompts in VS Code
- Git and Github login - these will make working with the workshop easier on you
- Azure subscription - in order to use Azure AI Document Intelligence and Azure AI Search you'll need a subscription (unless you are in the workshop on September 27, 2025 - see note below)
NOTE: For those of you in the workshop on September 27, 2025 - I will be providing you with the api keys to use predeployed Azure resources for the day only.
| Lab | Link to start page |
|---|---|
| Lab 0 - Getting started | Link |
| Lab 1 - Processing PDFs to be useful in RAG | Link |
| Lab 2 - Create your first agent | Link |
| Lab 3 - Refactor to use multiple agents | Link |
| Lab 4 - Doc Gen with a plugin | Link |
Jason is an independent Full Stack Solution Architect with a deep focus on Azure and Microsoft technologies. He is currently focused on helping customers integrate Gen AI functionality into their applications. He also helps run the Boston Azure AI user group, enjoys roasting his own coffee at home and runs ultra marathons every now and then.
LinkedIn | Twitter | Jason's Blog | Email
- azure-search-openai-demo is an open source RAG sample. A lot of the embedding service and saerch service logic was modeled after this project.
- Office-Word-MCP-Server is an open source MCP server for creating docx files. A lot of the document service logic is modeled after this project.