What it does

The Teams native ACD. A call queue holds callers with music, presents calls to a set of agents using one of four routing methods, and sends calls elsewhere when the queue is full, the wait is too long or nobody is available.

Key facts

• Routing methods: Attendant (ring everyone at once), Serial (fixed order, cycling), Round robin (even distribution), Longest idle (whoever has been Available longest). Microsoft recommends round robin or longest idle.
• Presence-based routing only offers calls to agents whose presence is Available. Longest idle switches it on automatically and you can't turn it off.
• Conference mode is on by default and recommended. It cuts connection time after the agent accepts, and it's required for consult/transfer, compliance recording, automatic recording and monitor/whisper/barge. Agents can hear up to two seconds of hold music on connect. Transfer mode is flagged as deprecated in H2 2026.
• Agent alert time: minimum 15 seconds, recommended at least 20. On Windows with a Bluetooth audio device the effective maximum ring is 120 seconds.
• Up to 200 agents per queue. Only 20 can be added as individual users; for more, use a group, a Teams channel or a Shifts schedule group (each up to 200). Groups take up to 8 hours to sync new members; channel-based queues can take up to 24 hours.
• Overflow: maximum calls waiting defaults to 50, range 0–200. Set it to 0 and the greeting can't play.
• Timeout: 0 seconds to 45 minutes.
• No-agents handling can apply to all calls or new calls only. With longest idle, "new calls only" doesn't work properly — a documented known issue.
• Exception destinations: disconnect, another queue, a person, shared voicemail, an auto attendant or an external number. PowerShell adds per-exception prompts (for example -OverflowRedirectPersonTextToSpeechPrompt, -TimeoutRedirectVoicemailAudioFilePrompt) on New-CsCallQueue/Set-CsCallQueue.
• Callback eligibility triggers: wait time exceeded, calls in queue reached, or calls-to-agent ratio exceeded. The caller's number must be publicly dialable E.164 and non-premium.
• Service level threshold: 0–2,400 seconds. Changing it mid-day skews that whole day's figure.

When to use / skip

Native queues are fine for internal service desks, reception overflow, small sales and support teams, and anywhere the requirement is "ring the right people, fairly, and do something sensible when they don't answer". Round robin plus presence-based routing plus conference mode covers most of that.

Skip native queues — or at least have the conversation — when you hear skills-based routing, priority by caller data, wrap-up time, multi-channel, workforce management or wallboards the supervisors actually trust. That's the native-vs-certified decision, which lives in the decision guides. Don't let it drift into "we'll just nest six queues" as a substitute for skills.

Configuration decisions

• Routing method: round robin for fairness, longest idle where agents also do other work and you don't want the person who just finished a call to get the next one. Attendant only for tiny teams or reception-style "anyone grab it".
• Agent source: security group for most queues (scales to 200, works with automatic recording); Teams channel only if the team genuinely lives in that channel and uses the Windows or Mac desktop client.
• Opt-out: allow agents to opt out or not. Supervisors usually want it on so agents doing callbacks or admin can step out.
• Overflow and timeout thresholds, and where they go. Voicemail-to-group is the usual default; a backup queue is better for revenue lines.
• Callback: whether to offer it and on which trigger. Make the timeout long enough for hold music to finish, the offer to play, and an agent to become free.
• PSTN connectivity of the resource account versus any external transfer targets — transfers between Calling Plan, Operator Connect and Direct Routing trunks aren't supported.

Gotchas

• Presence drives everything. An agent in a Teams meeting with "Busy" isn't Available, so a queue that looks staffed in the portal can hit no-agents handling at 10:00 on a Monday.
• Location-Based Routing and conference mode don't mix. With LBR on Direct Routing, agents get the toast but can't answer until someone else does or it times out.
• Private channels in channel-based queues send calls to all team members and assignment can randomly fail. Use standard channels.
• Skype for Business clients are excluded when presence-based routing is on, and longest idle doesn't work on them at all.
• Channel-based queues, monitor/whisper/barge and shared call history are desktop-only. Mobile and web agents don't get them.
• Callback, external overflow and timeout to an external number are outbound calls on behalf of the resource account. On Calling Plan numbers they need pay-as-you-go or Communications Credits since November 2025, or they fail.
• Individual-user assignment (max 20) isn't supported with automatic recording. Use a group from day one.

Consultant notes

• Set alert time to 20–30 seconds and make sure it's shorter than the agents' personal voicemail/forward timers, or calls end up in individual voicemail.
• Agree a queue naming standard and set the resource account display name to match — it's what agents see on the incoming call.
• Document every exception path on the call-flow diagram. The 18:00 "where did that call go" question is always an overflow or no-agents route nobody remembered.
• Prove conference mode, transfer and consult with the client's real handsets and headsets in UAT, not just the desktop client.
• If supervisors need real-time stats or to move agents in and out, that's the Queues app and Teams Premium — raise it early because it's a licence conversation.

---

Worth revisiting once transfer mode is formally retired — check whether any older queues are still on it