Chats, Channels, and What Your License Covers

Pull decisions and actions out of Teams chats and channels without inventing owners, rewrite messages without changing the facts, and learn which license tier turns each Copilot feature on.


What you'll learn

  • Choose the right Teams surface (chat, group chat, or channel) for a task
  • Extract decisions and actions using only explicit statements
  • Rewrite a message while preserving its request, deadline, and counts
  • Map a Teams AI feature to its tier and separate entitlement from condition
On this page

You're @mentioned in a channel with two hundred unread messages, and someone needs "the three decisions" in ten minutes. Copilot can summarize the thread and pull out actions and owners. It can also assign an item that nobody accepted. This lesson starts with the conversation itself, then moves to the license map so you can separate a weak summary from a feature you aren't entitled to use.

Pick the right surface first

Copilot works on the conversation you point it at, so the first move is choosing the right one:

What you need to work with Where to open Copilot
Messages in a one-to-one chat That chat
Messages in a team channel That channel
Messages in a multi-person group chat That group chat
Material from a different conversation Open that conversation first

Before you prompt, finish this sentence: "I need Copilot to work with the messages in ____." Then confirm the conversation you have open matches the blank: read its name and the first and latest relevant messages. Analyzing the wrong thread is the quietest way to get a confident, useless answer, and it's entirely avoidable with a five-second check.

Summarize a channel into four honest buckets· teams
Bad example

Summarize this channel and tell me who needs to do what.

Good example

Summarize this conversation in four sections: 1) confirmed facts, 2) decisions, 3) open questions, 4) action items. Use only explicit statements from the conversation. For each action item, include an owner and date only when stated. Write "Not stated" for missing fields. Don't convert a risk or a suggestion into an assigned action.

Why this works: A four-part summary where every action has a real owner or a visible "Not stated," and nobody's concern got silently promoted to a task.

Explicit means explicit

A trustworthy conversation summary keeps four categories apart: a fact someone reported, a decision the messages show was made, an action item someone explicitly committed to or was assigned, and an unknown where a required owner, date, or answer is absent. The word that does the work is explicit.

Marco writes, "the accessibility review found two issues." Ana writes, "I will fix the first issue by Friday." The thread supports one action: Ana's fix of the first issue by Friday. You can report the second issue as unresolved, but nobody has accepted it. A summary that says "both issues are owned" has created a commitment the conversation doesn't contain. Keep ability statements such as "I could look at that," suggestions, and concerns outside the assigned-action list. Before forwarding Copilot's summary, compare it with the messages that support each decision and action.

Rewrite without changing the ask

Teams can rewrite a message you're composing to fix the tone: turn a terse, frustrated line into something professional and collaborative. That's useful, and it's also where facts leak. Rewriting is allowed to change wording. The underlying request has to survive intact.

Compare the proposed rewrite with the original request, deadline, recipient, counts, and level of urgency. It must not add a promise, reason, or accusation. Take "need the revised budget by 3 PM. This has already slipped twice." The new version still needs 3 PM and "twice." If it says "soon," drops the count, or explains why the delay happened without a source, the friendlier tone has changed the message.

Rewrite a message without changing the ask· teams
Bad example

Make this message sound nicer.

Good example

Rewrite this to sound professional and firm: "Need the revised budget by 3 PM. This has already slipped twice." Preserve the 3 PM deadline and the fact that it has slipped twice. Use collaborative language, and do not add a reason or a new commitment.

Why this works: A calmer version that still carries the 3 PM deadline and the "twice" count, with no invented excuse attached.

Diff the rewrite before you send

Put the rewrite next to the original and check five things: same request, same deadline, same counts, same recipient, no new promise or reason. Tone is free to change. Everything on that list is not.

Agents answer within a boundary

A Copilot agent in a group chat is an AI participant scoped to a defined kind of question, say a purchasing-policy agent. The difference from a plain summary matters: a summary organizes a conversation, while an agent answers within an information boundary. A well-behaved agent distinguishes fields its available information answers from fields it doesn't, and it never fills the gap with something that merely sounds right.

Make the boundary explicit in your prompt and demand a specific phrase for gaps, such as "Not answered by the available policy." If you ask a policy agent for the required approver, the required documentation, and the escalation path, and the policy it holds covers only the first two, the correct answer flags the third as unanswered rather than inventing a plausible escalation chain. Never treat an agent as the final authority on a policy, an approval, or a commitment. Check the important fields against the governing source.

Ask an agent to stay inside its policy· teams
Bad example

Tell me how to get this purchase approved.

Good example

Using only the policy information available to you, identify the approval steps for a purchase above $10,000. Return the required approver, the required documentation, and the escalation path. If the available policy doesn't answer a field, write "Not answered by the available policy."

Why this works: An answer that fills only the fields the policy covers and flags the rest, so a plausible but invented escalation path can't slip through.

Which license turns this on?

Everything you've done in this module (live meeting questions, the recap features from Meetings: Recaps, Actions, and Follow-Ups, Channel Agents) sits on a licensing map with three tiers: baseline Teams AI (standard features, no Copilot license), Teams Premium (a paid add-on), and Microsoft 365 Copilot (a paid add-on). Both paid tiers cost money on top of your base plan. This course doesn't quote prices, because they change, and the durable knowledge is which tier turns on which feature.

Feature AI entitlement Extra condition
Suggested replies, noise suppression, voice isolation Baseline Teams AI An underlying Teams service
Base intelligent recap (AI notes, tasks, markers, chapters, topics) Teams Premium or Microsoft 365 Copilot None universal
Ask questions during a meeting Microsoft 365 Copilot None universal
Facilitator (real-time notes, agenda tracking) Microsoft 365 Copilot Facilitator added to the meeting
Channel Agent Microsoft 365 Copilot None universal
Multi-meeting audio recap Microsoft 365 Copilot Completed meetings available
Intelligent recap of an event Teams Premium Organizer, co-organizer, or presenter
Intelligent recap of a PSTN call Teams Premium or Microsoft 365 Copilot Teams Phone also required

Entitlement versus condition: the missing-button trap

Read that table as two separate questions. The AI-entitlement column answers "which tier gates this?" The extra-condition column answers "what else must be true?" They are not the same thing. A PSTN call recap needs a Teams Premium or Microsoft 365 Copilot entitlement and Teams Phone: one is the license gate, the other is a service requirement, and collapsing them produces wrong purchasing decisions. A PSTN call (to a regular phone number) carries the Teams Phone condition, while a one-to-one or group VoIP call between Teams participants does not, so classify the call before you decide.

A control you can't find proves only that the feature was not observed in that test. The client, a policy setting, the rollout state, or the communication type may explain it. Record the observation first. Then check entitlement, conditions, and configuration as separate questions. "I don't see the button" and "we didn't buy the license" require different evidence.

Not observed is not missing

A missing Copilot control means the feature wasn't observed, and any license conclusion needs separate evidence. Before you conclude "license missing," rule out client version, tenant policy, rollout state, and meeting or call type.

Try it yourself

Turn three team needs into a licensing map

For each need, name the exact feature, its AI entitlement, and any extra condition. Use the feature map above. No tenant access required.

  1. 01

    Project managers want AI-generated notes and recommended tasks after meetings. Name the exact feature, the AI entitlement, and any extra condition.

    Hint: This is base intelligent recap, and two tiers can satisfy it.

  2. 02

    Program leads want to ask Copilot questions live during meetings. Name the feature and the entitlement.

  3. 03

    The service desk wants intelligent recaps of calls made to regular phone numbers. Name the feature, the entitlement, and the extra condition, and say why a PSTN call differs from a VoIP one here.

A map that reads: base intelligent recap → Teams Premium or Microsoft 365 Copilot, no universal condition. Live meeting questions → Microsoft 365 Copilot. PSTN call recap → Teams Premium or Microsoft 365 Copilot, plus Teams Phone. If you wrote "relevant license" anywhere, you named a vibe instead of a tier, and that's the fix.

Key takeaways

  • Pick the conversation surface that holds your evidence (chat, group chat, or channel) before you open Copilot.
  • Report work as assigned only when a message contains an explicit commitment. Keep concerns and suggestions separate.
  • A rewrite may change tone but must preserve the request, the deadline, and every count.
  • An agent answers within its available source. Make it flag "Not answered by the available policy" and verify against the real policy.
  • Match each feature to its exact tier. A missing control means "not observed," not "license missing."

Check your understanding

  1. 1. A chat says "the accessibility review found two issues" and "Ana will fix the first issue by Friday." How many action items are explicitly supported?

  2. 2. A rewrite changes "I will fix the first accessibility issue by Friday" to "Ana will resolve both accessibility issues before Friday so launch can proceed." What went wrong?

  3. 3. A completed meeting shows ordinary materials, but no AI notes, tasks, or chapters appear. No license or configuration record is available. What can you conclude?

  4. 4. Which feature requires Teams Phone in addition to its AI entitlement?

  5. 5. A group-chat agent gives you a clean, confident escalation path for a purchase, but the policy it was given never mentions escalation. What should you do?

Frequently asked questions

Terms used in this lesson

baseline Teams AI
Standard Teams AI features (suggested replies, noise suppression, voice isolation, media quality) that don't require a Copilot license.
entitlement
The license gate tied to a capability (baseline Teams AI, Teams Premium, or Microsoft 365 Copilot), as opposed to what a specific user can access in a given meeting.
Channel Agent
A Microsoft 365 Copilot capability that lets an agent participate within a Teams channel.
PSTN call
A call that connects through the public telephone network, such as a call to a regular phone number, distinct from an internet-based VoIP call.

Further reading