Blog

Microsoft’s AI ecosystem is confusing.

Not because the tools are bad, but because the naming overlaps everywhere. Copilot, Microsoft 365 Copilot, Copilot Studio, Azure AI Foundry, Fabric, Power Apps, Agent 365, Agent Factory , they all sound related, and they are, but Microsoft’s documentation does not always make the relationship clear.

This is my simple mental model after researching the Microsoft AI offering.

1. Copilot: where users experience AI

Start here.

Copilot is the user-facing AI experience.

There are two main versions:

Copilot Consumer

This is the personal version of Copilot.

You use it through:

• web
• Edge
• Windows
• personal Microsoft account

Microsoft 365 Copilot

This is the work version.

It lives inside Microsoft 365 and connects to work tools like:

• chat
• Word
• Excel
• Outlook
• Teams
• files
• meetings
• emails
• tasks

My simple explanation:

Microsoft 365 Copilot is the AI interface for work.

It is where users chat, search, summarize, create, use agents, and work with company context.

2. Agents: how Microsoft lets you extend Copilot

Inside Microsoft 365 Copilot, you can create or use agents.

An agent is a focused AI assistant for a specific task.

The easiest way to understand agent creation is by complexity:

1. New Agent
   Simple, no-code, built inside Copilot
2. Microsoft Copilot Studio
   More advanced no-code / low-code agent builder
3. Microsoft 365 Agents SDK / Azure AI Foundry
   Code-based developer approach

That is the most useful ladder.

New Agent

This is the easiest option.

You click New Agent and create a basic agent without writing code.

Copilot Studio

This is deeper than New Agent.

It is still no-code / low-code, but gives more control. You can build, test, update, publish, and manage agents.

Agents can plug into:

• Microsoft 365 Copilot
• Teams
• websites
• custom apps

My simple explanation:

Copilot Studio is the serious no-code agent builder.

Microsoft 365 Agents SDK / Azure AI Foundry

This is the developer path.

Use this when you need code, custom logic, backend control, models, data connections, security, and more advanced AI systems.

My simple explanation:

Azure AI Foundry is the AI backend for developers.

3. Microsoft-built agents and Frontier features

Microsoft also has built-in or Microsoft-created agents.

Some are marked as Frontier, which basically means early, experimental, or beta-style features.

Example:

App Builder

App Builder is a Frontier agent that creates an app.

My simple explanation:

App Builder is an agent that helps create lightweight business apps.

It seems to combine ideas from:

• Copilot
• Copilot Studio
• Power Apps

You use it from Microsoft 365 Copilot, but the result feels like a lightweight Power Apps-style application.

4. Power Apps: internal business apps

Power Apps is Microsoft’s low-code tool for building internal business apps.

Examples:

• timesheet app
• IT support ticket app
• PTO request app
• inventory tracker
• approval workflow

My simple explanation:

Power Apps is for building internal business apps without traditional software development.

This is different from Copilot Studio.

Copilot Studio = build agents
Power Apps = build apps

But now AI is making these areas overlap.

That is part of the confusion.

5. Microsoft Fabric: business data

Microsoft Fabric is the data platform.

It is for storing, processing, analyzing, and connecting business data.

Related ideas:

• OneLake
• Power BI
• analytics
• Fabric IQ
• data agents
• business data for AI

My simple explanation:

Fabric is where business data can live so analytics and AI can use it.

Important note: Fabric does not magically contain all company data. It has to be enabled, set up, governed, and paid for.

6. Work IQ: work context

Work IQ is not really something I think of as a normal product.

It is more like Microsoft’s intelligence layer for work context.

It helps Copilot understand:

• emails
• files
• meetings
• chats
• tasks
• permissions
• tools
• organizational knowledge

My simple explanation:

Work IQ is the context layer behind Microsoft 365 Copilot.

I would not explain it as an app. I would explain it as the thing that helps Copilot understand work.

7. Agent 365: managing agents

Microsoft Agent 365 is for managing agents across an organization.

My simple explanation:

Agent 365 is the control plane for agents.

It is about:

• governance
• security
• discovery
• management
• compliance
• lifecycle
• shadow agents

You do not start here as a normal user. This becomes important when a company has many agents.

8. Agent Factory: enterprise agent program

Microsoft Agent Factory sounds like another product, but I think of it more as Microsoft’s enterprise program for scaling agents.

It connects many Microsoft AI pieces together:

• Microsoft 365 Copilot
• Copilot Studio
• Azure AI Foundry
• Fabric
• GitHub Copilot
• Azure AI Search
• governance tools

My simple explanation:

Agent Factory is Microsoft’s enterprise approach for helping companies build and scale agents.

It is not the same as Agent 365.

Agent 365 = govern agents
Agent Factory = scale agent adoption

Quick reference

Copilot
= consumer AI assistant
Microsoft 365 Copilot
= AI assistant for work
New Agent
= easiest no-code way to create an agent
Copilot Studio
= advanced no-code / low-code agent builder
Microsoft 365 Agents SDK
= developer SDK for custom agents
Azure AI Foundry
= AI backend platform for developers
Power Apps
= low-code business app builder
App Builder
= Copilot agent that creates lightweight apps
Microsoft Fabric
= business data and analytics platform
Work IQ
= work context layer behind Copilot
Agent 365
= governance and management for agents
Agent Factory
= enterprise program for scaling agent adoption

The shortest version

If you are new to Microsoft AI, think of it like this:

Use AI:
Microsoft 365 Copilot
Create a simple agent:
New Agent
Create a serious no-code agent:
Copilot Studio
Build custom AI with code:
Azure AI Foundry / Agents SDK
Build internal apps:
Power Apps / App Builder
Use business data:
Microsoft Fabric
Manage agents:
Agent 365
Scale agents across the company:
Agent Factory

This is the mental model that helped me understand the clutter.

Microsoft’s documentation explains the pieces, but not always the map. This is my map.

My 2018 Mac mini is starting to show its age.

It has served me well, especially with the i7 processor and 32 GB of RAM, but it is beginning to struggle with heavier workloads and too many browser tabs. The bigger issue, though, is compatibility. Many of the newer AI development tools, including apps like Codex, are moving away from Intel-based Macs and are no longer supported on older Intel chips.

So I decided it was time for an upgrade.

This is the new MacBook Pro with the M5 chip and 64 GB of memory. Apple has been positioning these machines as capable AI workstations, and I’m curious to see how well that claim holds up in real-world use.

I’m especially interested in experimenting with local AI tools such as Ollama, LM Studio, OpenCode, Open WebUI, and AnythingLLM. Based on my research, this machine should be able to run models such as Qwen3 30B, Gemma 31B, and possibly Llama 3.3 70B, depending on quantization, memory requirements, and performance expectations.

Over the next few weeks, I’ll be testing how practical this setup is for local AI development, coding assistance, and running large language models directly on the MacBook Pro.

I’ll post my findings as I go.

Hi everyone! My last WordPress site was hacked and I lost all my data, so I’m starting fresh.

I’ve learned the hard way that self-hosting isn’t always worth it. Keeping everything updated takes too much time, and it eventually became a security risk.

This time, I’m running my blog on WordPress.com. I needed something low-cost and fully managed so I don’t have to deal with security patches anymore. Let’s see how it goes!

Let’s see how it works out this time.