Every Microsoft Copilot for Business, Explained Without the Naming Headache

Microsoft named a lot of different things Copilot.

Some are licenses. Some are app features. Some are agents. Some are builders. Some are security tools. Some are old names that now point to newer names. Some are basically "Copilot, but inside the thing you were already using," which is honestly a lot of the problem.

So yes, this is annoying.

Here is the practical version: if you are trying to figure out which Microsoft Copilot your business actually needs, do not start with the product names. Start with the job.

How To Use This Guide

Do not treat this like homework.

The goal here is not to get you to read every word. It is to give you a north-star reference for Microsoft business Copilot decisions. Use the cheat sheet when you need the quick map. Use the navigation table when someone drops a product name into a meeting and you need the medium dive. Read the whole thing only if you really like Microsoft brand names.

The goal is to make the ecosystem easier to place in your head:

• Quick look: use the cheat sheet.
• Medium dive: jump to the section for the lane you care about.
• Planning mode: read the readiness, licensing, and rollout sections before assigning seats or building agents.

The Practical Answer

There is not one Microsoft Copilot for business. There are several Copilot lanes:

The mistake is trying to ask, "Which Copilot should we buy?" before asking, "What work are we trying to improve, what data does it depend on, and what risk is there?"

Microsoft Copilot Cheat Sheet

If you already understand Claude, Gemini, or ChatGPT this helps you place the Microsoft product in your head without pretending the products are identical.

One AI-speak translation before the table, because Microsoft says agent a lot: an agent is usually a saved AI helper with instructions, approved knowledge, and sometimes tools it can use. Sometimes it only answers questions. Sometimes it can draft, route, create, update, or trigger work. That last part is why agents need more governance than a normal chat prompt.

Where To Go From Here

Do not read this in order unless you want to. Pick the lane that matches the decision in front of you. Future you (and me) can come back and use the same table when Microsoft renames half of this again.

First: Pick The Lane

Most Copilot confusion comes from one word covering five different things.

When someone says "Copilot," ask which one of these they mean:

If you skip this step, you get the AI rollouts you hear about in the news: spend a lot of money on licenses, assign them to whoever asked first, run a demo, and then wonder why you're not getting good results for the money you burnt.

Part 1: Everyday Work Copilots

When most people say "Copilot for my business," they usually mean this lane first. At least at the beginning.

Microsoft 365 Copilot

Microsoft 365 Copilot is the paid work Copilot that sits across Microsoft 365 apps and uses organizational context through Microsoft Graph. Microsoft describes it as working with apps like Word, Excel, PowerPoint, Outlook, Teams, Loop, and others, and as using content the user has permission to access.

The important phrase is "the user has permission to access."

Copilot is not a secret company brain. It sees through the signed-in user's permissions and works inside the Microsoft 365 environment you already have. Good for data security. Less fun if your SharePoint and Teams permissions are a decade of "we'll clean that up later."

Where it actually helps:

• Drafting, rewriting, and summarizing documents.
• Summarizing email threads and drafting replies.
• Preparing for meetings from related content.
• Asking questions across accessible files, chats, meetings, and mail.
• Turning meeting transcripts into notes, decisions, and action items.
• Helping users work inside Word, Excel, PowerPoint, Outlook, Teams, Loop, OneNote, Whiteboard, and Forms.

• They assume it fixes content governance. It does not.
• They assign licenses before choosing high-value workflows.
• They treat meeting summaries gospel.
• They do not clean up SharePoint, Teams, OneDrive, or stale content.
• They forget Teams transcription and recording adoption control how useful meeting Copilot can be.
• They do not explain what users should do differently on Monday morning.

Microsoft 365 Copilot Chat

Microsoft 365 Copilot Chat is not the same thing as paid Microsoft 365 Copilot.

The difference is: Microsoft 365 Copilot uses organizational data and the web and requires an add-on license. Microsoft 365 Copilot Chat uses the web and lets users provide organizational data, and it does not require that additional Microsoft 365 Copilot license for eligible users.

That sounds like a small difference, it is not.

Copilot Chat is closer to the general chat experience people know from ChatGPT, Claude, or Gemini, but with Microsoft work identity, admin controls, and commercial protections. It can be a good starting point for secure AI usage, especially if the alternative is employees pasting company content into random consumer tools.

But do not confuse "included chat" with "full Microsoft 365 work-grounded Copilot." If someone expects the full experience across Outlook, Teams, Word, Excel, PowerPoint, Graph, and app context, that is Microsoft 365 Copilot, not Microsoft 365 Copilot Chat.

The sneakiest part is agents. Copilot Chat may be included for eligible users, but they can't build or use agents without additional costs, Copilot Studio capacity, and Azure subscription requirements.

The Microsoft 365 Copilot app

The Microsoft 365 Copilot app is Microsoft's front door for a lot of the work experience now.

It's designed to be the place that brings together Search, Chat, Agents, Pages, Notebooks, content creation, and the various Microsoft 365 Copilot app experiences.

For admins, the app turns Copilot from "one feature" into "a small ecosystem you have to manage":

• Which users have paid Microsoft 365 Copilot?
• Which users have Copilot Chat?
• Which agents can users create or use?
• Is web grounding enabled?
• Which apps are pinned?
• Which experiences are available in your cloud and region?
• Which data is searchable?
• Which interactions are retained, audited, or discoverable?

Copilot in the Microsoft 365 apps

• Word
• Excel
• PowerPoint
• Outlook
• Teams
• Loop
• OneNote
• Whiteboard
• Forms
• OneDrive
• SharePoint
• Clipchamp

Usually, treat these as app experiences under the Microsoft 365 Copilot umbrella, not as separate products to buy one by one. There are differences by license, tenant configuration, app version, feature maturity, cloud, and admin settings.

For a business leader, the useful question is not "Does Excel have Copilot?"

Better question: "Do we have a repeatable workflow where Copilot can save time or improve quality without creating a mess of new things to review?"

Same for Outlook. Same for Teams. Same for OneDrive. Same for SharePoint.Start with the work. Pick the Copilot surface. Then check whether the tenant can actually support it. In that order, even though everyone wants to skip to the fun part.

Copilot Search

Microsoft 365 Copilot Search is an AI search experience across Microsoft 365 and connected data sources.

Great when the content estate is clean. Less great when Copilot becomes the fastest way to discover content from 2013, poorly named files, ownerless sites, and permissions that made sense six years ago.

Search is only as good as the information estate behind it.

If you want Copilot Search to feel trustworthy, you need boring things:

• Content owners
• Sensitivity labels
• Permission cleanup
• Site lifecycle rules
• Authoritative source locations
• A plan for stale content
• A way to handle oversharing before users discover it with nicer wording

Pages, Notebooks, and Create

Microsoft 365 Copilot also includes newer work surfaces like Pages, Notebooks, and Create.

Think of these as AI-native workspaces and artifacts inside the Microsoft 365 Copilot world. They are similar in spirit to things like ChatGPT Canvas or Projects, Claude Artifacts or Projects, and Gemini Canvas or Notebook-style workflows.

The goal is not "a new version of Word" or "a new version of OneNote." It is a new work surface where Copilot can be more than a chat box: a place where prompts, files, outputs, collaboration, and work context live together.

Part 2: Agents And Governance

The first AI tool wave was mostly chatbots. You put in a request and got an answer back.

The next wave is agents.

Microsoft agents are focused assistants that can work with specific data sources, tasks, roles, or business processes. That sentence sounds simple. The annoying part is that "agent" can mean several different things depending on where it lives, what data it can see, and whether it can take action.

Microsoft 365 agents

Microsoft 365 agents are agents that live inside the Microsoft 365 Copilot ecosystem. Some are built by Microsoft. Some can be built by your organization. Some are role-based, some are process-based, some are grounded in Microsoft 365 data, and some connect to external systems. Same word, several shapes.

There are a few different agent shapes worth separating:

That split matters because "agent" does not always mean the same admin path, data path, cost model, or risk model.

Microsoft's prebuilt agents are useful less as a shopping list and more as a pattern:

• Researcher: an agent that can dig into a topic using Microsoft 365 data, public web data, or both, then produce a report.
• People Agent: an agent that helps you identify the right person inside the company based on skills, role, or context.
• Writing Coach-style helper: an agent pattern for improving drafts, sharpening tone, and helping users get better at writing over time.

• Naming agents: agents that know our naming conventions and help people name new resources consistently. Not glamorous. Very useful.
• Project Management Agent: an assistant that helps engineers turn technical work into project plans, then hand those plans to project managers without losing the plot. The best version speaks enough "engineer" for PMs and enough "PM" for engineers. It can also connect to the project management system so executives can pull client status updates without waiting on someone else to build a report.
• RFP Builder: an agent that takes in an RFP, reviews previous RFPs, and drafts a proposal using our language, pricing, and templates.
• Askio: a simple Microsoft 365 helper for quick "how do I do this?" questions in Word, Excel, PowerPoint, Outlook, Teams, and the rest of the usual Office suite.

That is the practical version of agents: a defined helper with a job, a boundary, and a reason to exist. Not "a bot that knows everything."

Good first targets:

• HR policy lookup with citations to approved sources.
• Sales account prep from CRM, email, meetings, and documents.
• IT help desk triage that creates tickets with the right fields.
• Project status synthesis from Teams, Planner, documents, and meetings.
• Finance reconciliation support with human approval.
• Customer service case research and summarization.

• "Make an agent that knows everything."
• "Make an agent that replaces the person who currently understands the process."
• "Make an agent that handles our whole sales pipeline."

Agents in SharePoint

Agents in SharePoint deserve their own mention because they are one of the more practical on-ramps.

The idea is simple: create an agent that answers questions from selected SharePoint content. That can be great for:

• HR policy sites.
• Safety procedures.
• Sales enablement libraries.
• Legal guidelines.
• New hire onboarding.
• Project document sets.
• Standard operating procedures.

SharePoint agents follow SharePoint permissions and organizational security policies. That helps. It also means the agent is only as trustworthy as the selected site's content, owners, permissions, lifecycle, and sources.

An agent over a clean policy site can be useful fast. An agent over a random shared site can become a headache-inducing waste of time.

Prebuilt Microsoft agents

Microsoft is also shipping more prebuilt agents across Microsoft 365 Copilot. Broadly, the pattern is still more "gather, synthesize, and draft" than "go run the whole business process," although that line is moving.

The useful thing is the direction: research, analysis, people, learning, workforce insights, workflows, sales, service, and employee self-service are all places Microsoft is trying to turn Copilot from a blank chat box into a more specialized helper.

Agent Builder

Agent Builder is the simpler way to create agents from inside Microsoft 365 Copilot.

Use it for narrower, lower-risk personal assistants where the scope is clear and the data is already reasonably controlled.

Do not use it as an excuse to avoid governance. Simple creation is useful. Simple creation at scale without ownership is how you get half the company using an agent linked to one person's account.

Microsoft Copilot Studio

Copilot Studio is Microsoft's more serious agent builder platform.

Think of Agent Builder as "I need a scoped helper." Think of Copilot Studio as "we need an agent that can be connected, published, governed, and supported."

Use it to build, customize, publish, and manage agents for a team or the whole organization. The simple version might be an in-tenant assistant that answers questions using approved Microsoft 365 content and supported third-party data. The more serious version might watch a customer service inbox, compare incoming requests against the knowledge base, draft a response or next steps, and route the work for review.These agents can be published across Microsoft 365, Teams, SharePoint, websites, apps, and other channels. Once you are here, you are doing platform work, which means the familiar Power Platform realities come with it:

• Environments
• DLP policies
• Connectors
• Authentication
• Dataverse
• Capacity
• ALM
• Maker permissions
• Publishing controls
• Monitoring
• Knowledge sources
• HTTP actions
• Skills
• Event triggers
• Channels

This is the part where "we just need a chatbot" becomes "who owns the environment, connector, data, deployment, and support model?"

Copilot Studio is powerful, but it is not magic. It is a platform. Platforms need owners, standards, intake, lifecycle management, and support paths.

Microsoft Agent 365

Microsoft Agent 365 is the governance/control-plane side of the agent story.

Microsoft positions it around observing, governing, and securing agents across the organization. That includes things like registry, agent analytics, onboarding, integration management, lifecycle management, audit and logging, data compliance, access control, data security, and threat protection.

Microsoft would not be building an agent control plane if agent sprawl were imaginary.

If every team can create or buy agents, IT and security need a way to answer basic questions:

• What agents exist?
• Who owns them?
• What data can they access?
• What tools can they use?
• What actions can they take?
• What logs exist?
• What happens when an owner leaves?
• What agents should be retired?
• Which agents are risky?

Agents change the risk model

A normal AI chat usually creates text.

An agent can combine instructions, data, tools, actions, connectors, permissions, memory or state, publishing channels, cost, and logs.

That means your governance model has to answer more than "can users prompt it?"It needs to answer:

• Who can create agents?
• Who approves agents?
• Who owns each agent?
• Which knowledge sources are allowed?
• Which connectors and HTTP actions are allowed?
• Can the agent be used by anonymous users, guests, customers, or only employees?
• What actions can it take without approval?
• What data can it retrieve?
• What logs exist?
• How is usage billed or metered?
• How do you retire stale agents?

Part 3: Business-Process Copilots

The business-process names are where Microsoft really starts testing your patience.

Microsoft has used names like Copilot for Sales, Sales in Microsoft 365 Copilot, Copilot in Dynamics 365 Sales, Sales agent in Microsoft 365 Copilot, Service in Microsoft 365 Copilot, Copilot for Service, finance agents, and Dynamics 365 agents. Some product pages have shifted. Some names are being folded into the broader Microsoft 365 Copilot agent story. And yes, the old blog posts are still out there confusing everyone.

My advice: Start with one of the general Copilot tools (Microsoft 365 Copilot, Microsoft 365 Copilot Chat) and see if that meets 80% of your needs, then check the current product page, licensing guide, and platform documentation for the specific workload.

Sales agent and Copilot in Dynamics 365 Sales

Copilot in Dynamics 365 Sales helps sellers summarize records, catch up on recent changes, prepare for meetings, get account news, use SharePoint content, and work with email.

The Sales agent in Microsoft 365 Copilot is the evolution of the sales Copilot experience in Microsoft Dynamics 365 Sales, and that Sales agent will become the preferred way to access Copilot assistance in Dynamics 365 Sales.

That is a sentence only a Microsoft ecosystem could produce, but the meaning is useful:Sales Copilot is moving toward an agent experience across Microsoft 365 apps AND Dynamics 365 Sales.

If you are planning a sales agent rollout, validate:

• Which CRM is in scope, such as Dynamics 365 Sales or Salesforce.
• Which Microsoft 365 apps sellers actually use.
• Which records and files sellers can access.
• Whether meeting prep, email follow-up, opportunity summaries, or account research are the first use cases.
• Whether your CRM data is clean enough to trust in summaries.
• Which license path applies now, not six months ago.
• Whether Teams meeting transcripts are enabled where meeting summaries matter.
• Whether Power Platform connectors and DLP policies allow the CRM integration.
• Whether the first user sign-in creates supporting Dataverse/Power Platform environment behavior you actually want.

Sales Qualification Agent

The Sales Qualification Agent is another example of where the Copilot ecosystem is pushing from "assistant" toward "workflow agent."

Note: It can help with lead research and qualification patterns, but it should not replace seller judgment, qualification rules, or CRM hygiene.

If lead data is weak, territory rules are unclear, or sales and marketing disagree on what qualified means, an agent will not settle the argument. It will just be authoritatively confused.

Service in Microsoft 365 Copilot and customer service agents

Service in Microsoft 365 Copilot is aimed at customer service representatives working across Outlook, Teams, CRM environments, and service workflows. Primarily for Dynamics 365 Customer Service and Salesforce, plus embedded widgets for non-Microsoft CRM systems.

Customer service agents can help with case management, intent, summarization, response drafting, and support workflows.

That can be genuinely useful because customer service work has repeatable patterns: cases, histories, knowledge articles, emails, chats, escalations, resolution notes, SLAs, and handoffs.

The risk comes from the same place.

If the knowledge base is stale, the agent will sound confident with stale information. If case routing is messy, the agent will inherit that mess. If approval rules are vague, agents will make the ambiguity faster.

Use customer service agents where the workflow is well-defined and the source material has owners. That sentence sounds obvious until you see how many support knowledge bases were all owned by Ted, who's been gone for 3 years.

Finance, supply-chain, field-service, and ERP agents

Microsoft is also pushing agents into finance and operations scenarios, such as account reconciliation and supplier communication.

These are not "cute productivity assistants." They sit near money, vendors, controls, approval chains, and audit.

Before you deploy finance or supply-chain agents, define:

• Human approval points
• Exception handling
• Audit requirements
• Data boundaries
• Who owns the process
• Which actions the agent can take
• What the agent is allowed to draft versus submit

The button exists. Great. That does not mean its okay to just drop it in.

For Dynamics 365 Customer Service, Field Service, Finance, Supply Chain Management, and Business Central, treat Copilot feature names and agent availability as release-plan dependent. Microsoft updates these workloads frequently.

Part 4: Data And Maker Copilots

Copilot in Power BI

Copilot in Power BI helps users ask questions about reports, summarize insights, create visuals, and help report authors build reports or work with DAX and semantic models.

This one is easier to place in your head than most of the Copilot naming mess. If you have used ChatGPT Advanced Data Analysis, Claude for data analysis, or Gemini with spreadsheet/data workflows, you get the general idea.

But Power BI Copilot is not just "upload a spreadsheet and ask questions." It depends on Power BI and Fabric concepts:

• Semantic models
• Reports
• Workspaces
• Apps
• Fabric capacity
• Admin settings
• User permissions
• Data preparation

An important licensing/admin point: Copilot in Power BI requires paid Fabric capacity. Pro or Premium Per User alone is not enough.

Also watch capacity. Copilot can consume Fabric capacity, so a broad rollout can create throttling or disruption for analytics workloads.

Copilot in Microsoft Fabric

Fabric Copilot extends AI help across Fabric workloads like Data Engineering, Data Science, Data Factory, Data Warehouse, Real-Time Intelligence, and Power BI.

For technical teams, Fabric Copilot can help with data pipelines, queries, summaries, and analysis. For leaders, the important part is simpler:

Fabric Copilot is only as useful as your data platform is usable. Not elegant, but accurate.

If your data estate has unclear ownership, duplicated models, no trust layer, and undocumented metric definitions, Copilot may help people move faster, but not necessarily in the same direction.

Also watch preview status, regional processing, sovereign-cloud support, and cross-geo settings. The Fabric story is powerful, but the admin evaluation is not optional.

Copilot in Power Apps

Copilot in Power Apps helps makers create and modify apps with AI assistance.

Useful for prototypes, internal tools, forms, and workflow apps. The catch is the same old Power Platform catch: someone still owns the environment, connectors, DLP, support, and app lifecycle.

Questions to ask:

• Which environment should this app live in?
• Who owns it after the maker moves on?
• What connectors does it use?
• Does DLP allow the data combination?
• Does the app need Dataverse?
• Is there an ALM path?
• How will support work?

Copilot in Power Automate

Copilot in Power Automate helps users create cloud flows from natural language and refine automations.

That is a good fit for repeatable workflows:

• Approval routing
• Notifications
• Intake processing
• Status updates
• Simple handoffs
• Scheduled reporting

It will also create fragile automation when the process is half-understood.

Before you automate, ask:

• What triggers the flow?
• What system is the source of truth?
• What happens when the flow fails?
• Who gets alerted?
• What data crosses boundaries?
• What approvals are real versus decorative?

Part 5: Security And IT Copilots

Security and IT Copilots are different from general productivity Copilots. They should feel different in the rollout plan too.

They operate closer to incidents, access, devices, compliance, cloud resources, and operational risk. They need tighter controls and clearer ownership.

Microsoft Security Copilot

Microsoft Security Copilot is built for security and IT workflows. Microsoft documents it across standalone and embedded experiences, including Defender, Entra, Intune, Purview, Sentinel, Defender for Cloud, threat intelligence, plugins, connectors, promptbooks, and agents.

The best analogy is not "ChatGPT for security."

The better analogy is "an AI analyst layer inside Microsoft security operations."

That matters because the value depends on your security tooling, telemetry, roles, plugins, promptbooks, and response process. If those are immature, Security Copilot can still help with summarization and investigation, but it will not magically create a mature SOC.

Good places to use it:

• Summarizing incidents
• Explaining scripts or suspicious activity
• Triage support
• Threat intelligence enrichment
• Promptbooks for repeatable investigation patterns
• Reporting and handoff notes
• Embedded assistance in Defender, Sentinel, Entra, Intune, and Purview workflows

• Role assignment
• Capacity and billing
• Security Compute Units and overage policy
• Plugin configuratio
• Audit logging
• Data sharing settings
• Government-cloud availability
• Over-trusting generated explanations
• Assuming "AI-assisted" means "approved"

Security Copilot can make good analysts faster. It can make weak process look polished too. Different problem.

Defender, Entra, Intune, Purview, Sentinel, And Other Embedded Surfaces

When you see Copilot inside Defender, Entra, Intune, Purview, Sentinel, or similar tools, treat it as a Security Copilot experience unless Microsoft is clearly selling something separate.

That framing keeps the map cleaner. Mostly.

You do not need a separate strategy for "Entra Copilot" and "Purview Copilot" as if they are unrelated products. You need a security and compliance AI strategy that says:

• Which roles can use these features?
• Which data can be accessed?
• Which prompts and responses are logged?
• Which outputs need human review?
• Which actions require approval?
• Which use cases are approved first?

Azure Copilot

Azure Copilot is Microsoft's AI assistant for Azure. It can help with resource questions, design, troubleshooting, operations, cost, and portal workflows.

The closest general analogy is asking ChatGPT, Claude, or Gemini for Azure guidance, except Azure Copilot can use Azure context the signed-in user has permission to access.

That is the upside. The catch is the usual Azure catch: permissions and operational guardrails decide how far you should let it go.

Good places to use it:

• Finding resources
• Understanding configuration
• Troubleshooting
• Cost and optimization questions
• Drafting deployment or architecture guidance
• Explaining what is running in an environment

• Preview features
• Regional availability
• Tenant and subscription access
• Action confirmation
• RBAC boundaries
• Privileged Identity Management
• Azure Policy
• Resource locks
• National cloud availability
• Network access to required services
• Overreliance on generated recommendations without engineering review

Part 6: Developer Copilot

GitHub Copilot is the developer Copilot.

This comparison is cleaner than most of the Microsoft AI pile, thankfully. The closest analogies are Claude Code, OpenAI Codex, and Gemini Code Assist.

GitHub Copilot can help with:

• Code completions
• IDE chat
• Repository-aware questions
• Pull request summaries
• Code review
• CLI assistance
• Agentic development tasks
• Documentation
• Tests
• Enterprise policy and governance

For businesses, focus on GitHub Copilot Business and GitHub Copilot Enterprise rather than personal developer plans.

The governance questions are different from Microsoft 365 Copilot:

• Which repositories can Copilot access?
• Are content exclusion rules needed?
• Is public code matching enabled or disabled?
• Which models are allowed?
• Are agentic features enabled?
• How are generated changes reviewed?
• What happens in regulated codebases?
• How are usage and quality measured?
• What is the policy for secrets, proprietary code, and dependency suggestions?
• Are network allowlists required?
• Are audit logs and agentic activity reports being reviewed?

Part 7: Edge Cases and Confusing Names

Industry-specific Copilots

Microsoft has industry-specific Copilot products and branded experiences, such as healthcare-oriented products like Microsoft Dragon Copilot.

If you are in healthcare, public sector, finance, manufacturing, or another regulated industry, validate:

• Product availability
• Compliance posture
• Data handling
• Licensing
• Integration model
• Regional availability
• Whether it is Microsoft 365, Dynamics, Nuance, Azure, or another Microsoft product family underneath

Education and government

Education and government cloud availability often varies by product and feature.

The names can look similar, but availability, licensing, compliance, functionality, and rollout paths may be different. This is especially true for GCC, GCC High, DoD, sovereign clouds, and regional data boundaries.

Windows, Edge, Copilot Pro, and consumer products

Windows Copilot, Edge Copilot, Copilot Pro, Microsoft 365 Personal/Family Copilot experiences, and Xbox Copilot are (hopefully) not the part of a business Copilot strategy.

They will confuse users because the logo and name look familiar (Thanks Microsoft...). But they are not the same tooling as the Microsoft Copilot Ecosystem for businesses.

Part 8: Licensing Quick Reference

Below is a quick reference for how you actually pay for these things. Check the live pricing pages for the actual costs.

Part 9: Readiness Before Licenses

Before you assign Microsoft 365 Copilot licenses broadly, check the boring things.

Part 10: Choosing The Right Microsoft Copilot

Use this decision flow.

The Bottom Line

• Microsoft 365 Copilot for everyday work.
• Copilot Chat for secure work chat.
• The Microsoft 365 Copilot app for Chat, Search, Agents, Pages, Notebooks, Create, and app access.
• App experiences inside Word, Excel, PowerPoint, Outlook, Teams, Loop, OneNote, Whiteboard, Forms, OneDrive, SharePoint, and Clipchamp.
• Microsoft 365 agents, SharePoint agents, prebuilt agents, and custom agents for specific tasks and workflows.
• Agent Builder and Copilot Studio for creating agents.
• Agent 365 for agent governance.
• Sales, Service, and Dynamics agents for sales, service, finance, field service, supply chain, and operations.
• Power BI and Fabric Copilot for data and analytics.
• Power Apps and Power Automate Copilot for makers and automation.
• Security Copilot for security and IT.
• Azure Copilot for cloud operations.
• GitHub Copilot for developers.

The useful question is not "Which Copilot is best?"

The useful question is: What work are we trying to improve, what data does it depend on, what is the risk, and which Copilot lane actually matches that job?