Chat vs Agents: Two Ways to Work
Copilot Chat is your general-purpose assistant. Agents are specialists. Learn the difference, when to use each, and why you might want to create your own agent.
Two ways to use AI at work
Think of it like phoning a general helpline vs calling a specialist.
Copilot Chat is the general helpline. You can ask it anything — summarise emails, draft a document, find information. It’s flexible, smart, and knows about all your work data. But it doesn’t have deep expertise in any one area.
An agent is a specialist. It’s been trained on specific knowledge (like your company’s HR policies or product catalogue) and given specific instructions (like “always respond in a friendly tone and only answer questions about employee benefits”). It does one thing really well.
When to create your own agent: When you find yourself asking Copilot the same type of question over and over, or when a team needs consistent, specialised answers from a specific set of documents.
Chat vs Agent — side by side
| Feature | Copilot Chat | Agent |
|---|---|---|
| Purpose | General-purpose assistant | Specialist for a specific task or topic |
| Knowledge | All your M365 data + web (if enabled) | Specific sources you configure (SharePoint sites, files, web URLs) |
| Instructions | None — responds to whatever you ask | Custom rules (tone, scope, limitations) |
| Consistency | Different users may get different styles | Same behaviour every time, for everyone |
| Setup required | None — just start typing | Create and configure (no coding needed) |
| Sharing | Each person uses their own Copilot | Share one agent with your whole team |
When to use each
Use Copilot Chat when…
- You have a one-off question across your work data
- You need to find something that could be in emails, chats, or files
- You want to draft something from scratch
- You’re exploring — not sure exactly what you need yet
Use an agent when…
- You have a repeating task that many people do (new hire questions, product FAQs)
- You need consistent answers from a specific set of documents
- You want to share a customised AI experience with your team
- You want to control the scope — keep the AI focused on one topic
Real-world: Dana's onboarding dilemma
Dana at Oakfield Healthcare gets 30+ questions a week from new hires: “Where’s the dress code policy?” “How do I book annual leave?” “What’s the Wi-Fi password?”
She tried using Copilot Chat, but each new hire had to formulate their own questions and sometimes got answers from the wrong documents.
Solution: Dana created an Onboarding Agent — scoped to the HR SharePoint site, with instructions like “Only answer questions about Oakfield employee policies. Be friendly and concise. If you’re not sure, direct them to HR.”
Now every new hire gets the same consistent, accurate experience. Sam (the training coordinator) loves it because it freed up 3 hours a week of answering repetitive questions.
Why create your own agent?
Here are the most common reasons businesses build agents:
| Reason | Example |
|---|---|
| Reduce repetitive questions | HR onboarding bot answers the same 50 questions |
| Standardise team responses | Sales team uses an agent to draft consistent proposal emails |
| Scope to specific data | Legal agent only references approved policy documents |
| Save setup time | Instead of explaining context every prompt, the agent already knows |
| Share expertise | One person builds it, the whole team benefits |
Exam tip: The exam often presents scenarios where someone is using Copilot Chat for a task that would be better served by an agent (or vice versa). The key test: is this a one-off question (Chat) or a repeating need for consistent, scoped answers (Agent)?
Built-in agents you should know
Microsoft includes several pre-built agents and templates:
Pre-built agents (ready to use from the Agent Store):
- Researcher — helps you research topics in depth using web and work data (requires M365 Copilot license)
- Analyst — focuses on data analysis and visualisation (requires M365 Copilot license)
Agent templates (starting points for building your own):
- Prompt Coach — template for building a prompting assistance agent
- Writing Coach — template for building a writing improvement agent
- Meeting Coach — template for building a meeting preparation agent
You’ll learn more about finding agents in the Agent Store and building your own in Domain 2.
🎬 Video walkthrough
🎬 Video coming soon
Chat vs Agents — AB-730 Module 4
Chat vs Agents — AB-730 Module 4
~8 minFlashcards
Knowledge Check
Jordan at Peak Solutions notices his sales reps ask the same Copilot questions every week: 'What's our pricing for the Enterprise tier?' 'What are our differentiators vs Competitor X?' Each rep gets slightly different answers because they phrase questions differently. What should Jordan do?
Ava needs to quickly find out what was discussed about the Q4 budget in yesterday's Teams meeting. She's never asked this kind of question before and doesn't expect to again. Which is the BETTER choice?
Which of the following is a characteristic of a Copilot agent? (Select TWO)
Next up: Copilot works with your data — but how does it keep that data safe? Learn about privacy, security, and how data protection shapes what Copilot can tell you.