Microsoft Copilot AI is Taking the Lead
Copilot AI stole the show at Microsoft Build this week, as Satya Nadella shared a forward-thinking vision for embedding AI deeper into business systems (watch on YouTube).
These advancements are undeniably groundbreaking. But many businesses are struggling to keep pace with the rapid evolution of technology. Especially in the Microsoft ecosystem, new terminology is introduced constantly. Instead of simplifying things, it often adds to the complexity and doesn’t let people catch up.
In this article, we share our latest research on Microsoft’s evolving Copilot strategy and help you decode what it all means for your organization.
🔥 What’s new (Build 2025):
Copilot Tuning – low-code way to make agents “talk/think/work like you.”
Multi-Agent Orchestration – agents can now delegate to other agents
Copilot Chat, Search, Notebooks & Create – new canvases to reuse enterprise data.
Quick Clarification: How Is Microsoft Copilot Different from ChatGPT?
While some organizations are well underway in their Copilot adoption, we still hear this question from those who haven’t fully bought in yet:
“Why not just use ChatGPT for business AI?”
It’s a reasonable question—and one worth answering clearly.
The truth is, ChatGPT is great for general tasks like drafting content, summarizing documents, or answering open-ended questions. But when it comes to business-critical use cases, there are key differences that make Microsoft Copilot the better choice for organizations:
-
Built for secure enterprise use
Copilot runs entirely within your Microsoft 365 tenant. That means no data leaves your environment, and your content is never used to train public models—unlike ChatGPT, which stores and may learn from prompts unless you’ve explicitly disabled that setting. -
Connects directly to your business data
Copilot works natively with Microsoft Graph, Dataverse, SharePoint, and more. You’re not pasting in content or exporting files—it already has secure access to the context it needs. -
Understands your role and workflows
Because Copilot knows who you are in your organization, what you’re working on, and what tools you’re using, it can offer tailored, actionable suggestions—not just generic replies. -
Easy to extend, securely
With Copilot Studio, teams can build, customize, and control AI agents using low-code tools—while enforcing your org’s policies and data boundaries.
Let's get into Microsoft Copilot AI
At Microsoft Build, CTO Kevin Scott presented a diagram like the one below, highlighting Microsoft’s vision for the Agentic Web (with my own addition of the Dynamics 365 Copilots).
In this ecosystem, “Copilot” is best understood as the primary UI and branding for Microsoft’s AI stack. During his session, Kevin Scott provided a simple but powerful definition of an agent as:
“Something a human can delegate a task to.”
He expanded on that vision with the concept of the Agentic Web—an evolving ecosystem in which agents can also delegate tasks to other agents (watch the session on YouTube).
It may sound like another marketing term to have to remember, but it is well suited for qualifying this ecosystem that’s certainly been lacking in structure.
There are volumes of information related to this concept, but for this article we’re going to focus on the Copilot tooling that provides value to all business and technical Microsoft users.
Copilot for Everyone (Free Tier)
The free Copilot at copilot.microsoft.com is the same LLM-powered assistant embedded in Windows 11, Edge, Bing and Office mobile apps. Feature parity with ChatGPT4-o isn’t fully there yet, but Microsoft keeps closing the gap with faster model refreshes and first-party data integration.
Microsoft 365 Copilot
Microsoft 365 (M365) Copilot is Microsoft’s flagship enterprise AI assistant, designed to bring the power of large language models (LLMs) directly into the app’s employees use every day (Outlook, Teams, Excel, etc). Built on Microsoft Graph and deeply integrated with SharePoint, Dataverse, and OneDrive, M365 Copilot understands your organization’s data, context, and permissions, making it far more than just a ChatGPT wrapper.
At Build 2025, Microsoft introduced several new features that dramatically expand the reach and intelligence of M365 Copilot:
-
Copilot Chat – Embed Copilot into other apps and interfaces beyond Microsoft 365, extending its reach into custom business apps and workflows.
-
Copilot Search – Leverage Graph connectors to unify search across Microsoft data and external platforms like Salesforce, ServiceNow, and SAP.
-
Copilot Notebooks – Collect and work with diverse content—emails, documents, recordings, chats—within a structured workspace for long-form reasoning.
-
Copilot Create – Generate content including Office documents, images, and more using both enterprise data and real-time web input.
-
New Agents: Analyst & Researcher
-
Analyst specializes in data analysis tasks—synthesizing trends from Excel, creating visuals, and interpreting metrics.
-
Researcher performs deep synthesis on large, cross-source datasets to formulate hypotheses, summaries, and next steps.
-
Copilot Tuning (Preview): fine-tune models to use your domain language, KPI definitions and security boundaries—all inside your tenant. (Microsoft Learn)
Build Agents with Copilot Studio
The last key piece of this that isn’t directly in the prior diagram is Copilot Studio. Copilot Studio lets users build and deploy their own agents—for personal, team, or enterprise use cases. Though you can create agents in Azure and other means, Copilot Studio is meant for anyone to develop Agents with no/low-code type interactions.
Users can build agents using decision-tree logic (e.g., if user says ‘x’, respond with ‘y’). Or they can use generative AI to let the agent reason independently. There is also the ability to use Pro-Code based approaches to developing agents using Azure AI Foundry (AI Studio), but we’ll dive deeper into that comparison in another post.
Key Features
- Knowledge – Define what data sources the agent has access to such as SharePoint, Dataverse, and even external systems like DocuSign, SAP, etc.
- Tools – You can define tools such as Dataverse actions, 3rd party services, and even other Agents (using MCP protocol) to allow your agent to take action for users.
- Prompt by AI Builder – Copilot can integrate AI Builder prompts (Microsoft Learn) that enables your Agent to perform actions. These prompts utilize GPT models provided by Azure OpenAI Services behind the scenes. This enables better delegation and coordination of your agents / AI tooling.
Power Automate Agent Flows
One of the last item’s well call out from Build was the roll-out of Agent Flows in Copilot Studio (Microsoft Learn). Similar to Cloud Flows, however these have the added capability to allow users to trigger flows from Copilot and incorporate Natural Language into the workflow. These workflows run under your Copilot capacity allotments, and you can convert existing cloud flows to be agent flows as well. This enables workflow automation that wouldn’t be possible without practicing more complex logic, which means business users have more power in their hands than ever before.
Check out Kumo's Demo Copilot
Frequently Asked Questions (FAQ)
ChatGPT is a general-purpose language model that works with public internet data. Microsoft Copilot integrates directly with your business data in Microsoft 365, understands your organization’s context, and enforces enterprise security and compliance.
No. Microsoft 365 Copilot requires a $30 per user/month add-on license on top of a Microsoft 365 E3 or E5 subscription. In contrast, the free Copilot (copilot.microsoft.com) provides general AI assistance but doesn’t connect to your organization’s data, files, or tools.
Not necessarily. With Copilot Studio, business users can build and deploy agents using low-code tools and guided AI experiences. Developers can enhance those agents further using Power Platform, Azure, or custom connectors.
Ready to Explore your own Copilot Use-Case?
Here at Kumo, we work with Clients everyday to better understand the technology that is available to them and work rigorously to parse out the value from the fluff. If you’re looking for a partner that will work with you and put your needs first, you’ve come to the right place.
0 Comments