Microsoft IQ Series: a video-and-notebook course for Foundry IQ and Work IQ

What the IQ Series actually is, and who it is written for

This repository is a curriculum. The README describes it as a hands-on learning experience for Microsoft IQ, which it defines as a unified intelligence layer spanning three services: Foundry IQ, a managed knowledge layer connecting structured and unstructured data across Azure, SharePoint, OneLake and the web with permission-aware retrieval; Work IQ, which the README frames as organizational context, relationships and work patterns exposed to agents; and Fabric IQ, which unifies business semantics across data, models and systems. The stated audience is anyone building agents on those services who wants to see how the pieces fit before writing code. That is a narrower audience than the topic list suggests. If you are evaluating retrieval stacks generally, or you run on AWS or GCP, the material has nothing to say to you. The repository exists because the three IQs are separate products with separate entry points, and someone has to show how a knowledge source in Foundry IQ relates to an agent that also calls Work IQ tools. The episodes are the argument; the notebooks are the proof.

The episode structure and what each folder contains

Each episode follows a fixed three-part format: an introduction of roughly one minute described as an executive speech, a tech talk of about fifteen minutes with the product group and advocacy teams, and a one-minute close-out described as a doodle summary. That format tells you what the videos are for. They are orientation, not reference. The durable part of each episode is the folder beside it. Foundry IQ episodes ship a cookbook directory; Work IQ episodes ship a cookbook directory containing a README that points at lab instructions. The README notes that episode folders include either Jupyter notebook cookbooks or markdown lab instructions. That distinction matters when you plan time: a notebook you execute, a lab you follow. The repository is overwhelmingly Jupyter Notebook by language, which is consistent with a teaching artifact where the deliverable is a runnable cell rather than an importable module.

The Foundry IQ track, episode by episode

Three Foundry IQ episodes are listed. Episode one, Unlocking Knowledge for Agents, covers Foundry IQ's core components and where it sits in an agent architecture. Episode two, Building the Data Pipeline with Knowledge Sources, covers how different content enters Foundry IQ from various sources. Episode three, Querying the Multi-Source AI Knowledge Bases, goes into knowledge bases and multi-source query paths. Read as a sequence, the track moves from architecture to ingestion to retrieval, which is the order you would want if you were about to build something. The README says Foundry IQ episodes premiere every Wednesday at 9 AM PT starting 18 March 2026 on Microsoft Developer YouTube. Note the date. As of the repository's last push on 3 August 2026 the run has started, but anyone reading this earlier would find links to episodes that had not aired. The README also offers a badge: complete all three Foundry IQ cookbooks, fork the repository, save your notebook outputs, capture a final output screenshot per episode, open a badge request issue and complete the badge form. Badges are issued by the Global AI Community, so the README asks you to have an account first.

The Work IQ track and its protocol emphasis

The Work IQ episodes take a different angle. Episode one, Data, context, and tools at scale, introduces the Work IQ architecture and what the README calls protocol strategy across REST, A2A and MCP for enterprise agent experiences. Episode two covers A2A specifically: agent discovery patterns and prototyping for context-aware workflows. Episode three covers tooling with MCP and Copilot CLI, and the README describes it as exploring the unified MCP server of Work IQ and how to use it in Work IQ CLI and GitHub Copilot CLI. That third episode is the most concrete of the six, because it names a server and two clients. If you only watch one, watch that one. Work IQ episodes premiere at 9 AM PT on 2 June 2026. The badge path differs slightly from Foundry IQ: complete all three labs, fork the repository, save lab outputs as screenshots per episode, submit a badge request issue and complete the badge form. The README points you at the badge request issue form for the exact screenshots required, which means the requirements are not fully enumerated in the README itself.

Getting a cookbook running: what the README does and does not give you

The README gives you links, not setup instructions. Each episode entry points to a cookbook directory, for example ./Foundry-IQ/1-Foundry-IQ-Unlocking-Knowledge-for-Agents/cookbook/, and the Work IQ entries point to a cookbook README such as ./Work-IQ/3-Work-IQ-Tooling-with-MCP-and-Copilot-CLI/cookbook/README.md. The repository description mentions Azure deployment templates alongside the notebooks, so infrastructure provisioning is expected to live in those templates rather than in the top-level README. Beyond that, the top-level document does not list environment variables, Python version, package requirements, or authentication steps. Those have to come from the individual cookbook and lab READMEs. This is a real friction point and worth stating plainly: you cannot judge the setup cost of the IQ Series from the repository root. You have to open the cookbook for the episode you care about and read its prerequisites. The upside of that layout is that each episode is self-contained; the downside is that there is no single install path and no shared requirements file described at the top level.

Where this is the wrong tool

Three limitations are visible from the material. First, it is incomplete by design. The README states that the series kicks off with Foundry IQ episodes followed by Work IQ episodes, and that Fabric IQ content is coming soon. The Fabric IQ row in the episode table has no description, no video link and no cookbook. If your interest is business semantics across data and models, the repository currently offers a definition and nothing runnable. Second, the schedule is a dependency. Episodes premiere on fixed dates, so the material you can work through is bounded by what has aired. A reader who arrives before 2 June 2026 gets three Foundry IQ episodes and no Work IQ labs. Third, this is sample code with an MIT licence and no releases retrieved. There is no versioned artifact to pin, no changelog to consult, and no compatibility guarantee between a notebook and the current state of the underlying Azure services. Notebooks that call managed services age. If Foundry IQ's knowledge source API changes, the cookbook cells change with it, and you will not get a deprecation notice from this repository.

What to use instead if you need a library rather than a course

The honest alternative is the SDKs and frameworks you would otherwise build on. If you want retrieval over your own documents with a stable, versioned interface, something like LlamaIndex or LangChain gives you a package you install and pin, with a release history and an API contract, rather than a notebook you fork. The difference in approach is fundamental: those projects abstract over many vector stores and model providers and let you swap components, while the IQ Series teaches one vendor's managed services end to end and assumes you have bought into them. The IQ Series wins on fidelity. It shows the actual permission-aware retrieval path through Foundry IQ and the actual MCP server for Work IQ, which no general framework can demonstrate because those are Microsoft-specific services. It loses on portability and on the ability to freeze a working version. If your goal is to understand the Microsoft stack, a general framework will teach you the wrong abstractions. If your goal is to ship retrieval that runs anywhere, the cookbooks will not get you there.

Licence, maintenance and what the repository costs you over time

The licence is MIT, which permits commercial use, modification and redistribution provided the copyright notice and permission notice are retained. That covers the notebooks and templates in the repository. It does not cover the Azure services the notebooks call, which are billed separately and governed by their own terms, and the README does not address that boundary. This is not legal advice; check the licence file and your own agreements. On maintenance, the material is thin. There are no releases retrieved, so there is nothing to upgrade to and nothing to pin. The last push recorded is 3 August 2026, which tells you the repository is active but not how often it changes or whether earlier notebooks are kept in sync with later service changes. Practical cost is therefore not upgrade work but re-validation: every time you run a cookbook against live Foundry IQ or Work IQ, you are testing whether the notebook still matches the service. Budget for that, and treat the fork-and-screenshot badge flow as what it is, a completion record rather than a support channel.

Editorial conclusion

Adopt the IQ Series if you are already committed to Azure, Microsoft Foundry, Fabric and Microsoft 365 Copilot and want a guided path through Foundry IQ knowledge sources and the Work IQ MCP server before writing your own integration. Skip it if you need a production library, a stable API surface, or anything that works outside Microsoft's cloud: this is a curriculum with a March 2026 premiere date and a Fabric IQ track that has not shipped. Verify three things first: that you can meet the Azure prerequisites each cookbook lists, that the Work IQ MCP server is available in your tenant, and that the episode you need is actually published, since the Foundry IQ run starts 18 March 2026 and Work IQ starts 2 June 2026.

Official sources

1. Issues
2. License: MIT
3. microsoft/iq-series on GitHub
4. Project website
5. README