Chat concurrency· Support operations
Chat Concurrency — Why One Agent Can Hold Several Chats, and What It Costs
Quick answer: Chat concurrency is how many conversations an agent works at once. A phone agent takes one call at a time; a live chat agent can answer one customer while another is typing. Help desks call the cap a chat limit, capacity or assignment limit, and stop sending new chats to an agent who has reached it. A higher limit lets the same team take more chats, but each chat tends to take longer, so the gain is smaller than the limit suggests. Most teams set it per agent or per channel and test it against their own handle time and response times.
Limit versus average: two different numbers
| Number | What it is | Who sets it |
|---|---|---|
| Concurrency limit | The most chats the system will assign to one agent at once | An admin, in the help desk settings |
| Peak concurrency | The most chats an agent actually held at one moment | Happens; seen in reports |
| Average concurrency | How many chats an agent worked on, on average, while available | Happens; calculated |
The limit is a ceiling, not a description of the work. Call Centre Helper makes this point when it adjusts staffing formulas: because the number of chats an agent holds goes up and down, the factor it uses for two chats is 1.7, not 2, and for three chats it is 2.5, not 3.
A simple way to measure average concurrency, used here as our own working method: add up the minutes each chat was assigned to an agent, and divide by the minutes that agent was available for chat. Four hours available with 408 chat-minutes handled gives 408 ÷ 240 = 1.7.
Why handle time rises as concurrency rises
An agent holding three chats switches between them, reads back to catch up, and makes each customer wait while the others are answered. So the time each chat stays open grows. Call Centre Helper puts it plainly: "Because AHT goes up as you increase concurrency you start to lose any real benefit." That is the central trade-off, and it shows up in average handle time and in the gaps between replies that customers notice.
An invented example of the arithmetic. One agent's throughput is average concurrency divided by the time each chat stays open:
| Setting | Average concurrency | Minutes per chat | Chats per agent-hour |
|---|---|---|---|
| Limit 2 | 1.7 | 12 | 1.7 × 60 ÷ 12 = 8.5 |
| Limit 3 | 2.4 | 16 | 2.4 × 60 ÷ 16 = 9.0 |
Raising the limit by one adds about 6 percent more chats per hour in this example, while every customer's chat lasts a third longer. With other numbers the gain could be larger or could vanish, which is why a limit is something to test rather than copy.
How help desks set and count the limit
Each tool uses its own name and counts different things. As documented on 17 November 2026:
| Tool | Name of the setting | Scope | What the documentation says |
|---|---|---|---|
| Zendesk | Capacity rules (omnichannel routing) | One rule per agent; a default rule covers new agents | Messaging capacity up to 20, applied to messaging conversations and live chats separately; quiet messaging conversations count only if an admin turns that on, which is off by default |
| Intercom | Assignment limit (balanced assignment) | Workspace, per teammate, and per team inbox | Limits apply across the whole workspace; an inbox limit is checked on top of the teammate limit; balanced assignment is on the Expert plan |
| HubSpot | Chat capacity limit | Account-wide default | Default 3 active chats, maximum 99; Service Hub Enterprise |
| LiveChat | Chat limit | Per agent, set in the Team section | When every accepting agent is at the limit, new visitors go into the queue |
Two details matter more than the number itself. First, manual assignment usually bypasses the limit: Zendesk and Intercom both note that assigning a chat by hand can push an agent past it, and HubSpot lets chats be assigned by hand when every agent is full. Second, what counts towards the limit differs: Intercom counts open conversations across all inboxes, while Zendesk leaves inactive messaging conversations out unless you change the setting. Compare these settings in our Intercom review, and see Zendesk, HubSpot Service Hub and LiveChat.
What happens when every agent is full
New chats wait. LiveChat puts visitors in a chat queue and can show their position and estimated wait. HubSpot leaves new chats unassigned and lets each chat channel choose what visitors see at maximum capacity: a wait message, a hidden launcher, or no change. Intercom holds conversations in the team inbox until a teammate has room. In every case the waiting time lands on your first response time and on your service level agreement, so the limit and the SLA have to be set together.
Concurrency and staffing
Staffing formulas were built for one contact per agent. Erlang C is usually adjusted for chat by dividing handle time by a concurrency factor, or by dividing the agents it returns by average concurrency. Both use the average, not the limit. If you plug the limit into a staffing calculation, you will plan too few people. The full method is in the guide to forecasting live chat staffing.
Where chatbots and AI fit
A bot in front of the team does not raise an agent's concurrency, but it changes the chats that reach agents. When a bot collects the order number and the problem before the human handoff, an agent can pick up a chat faster, which can make a higher limit workable. When the bot already answered the easy questions, the chats left are harder, and a lower limit may suit them better. Agent assist tools that suggest replies are another way to keep handle time down at higher concurrency; check the effect in your own reports before raising the limit.
FAQ
What is chat concurrency? The number of live chat or messaging conversations one agent handles at the same time. It is the main way chat differs from phone support, where an agent takes one call at a time.
How many chats can one agent handle at once? It depends on how complex the chats are and how fast customers reply. HubSpot's default limit is 3; Call Centre Helper gives 3 or 4 as an example. None of the sources we read sets a standard, so measure handle time and response times as you change the limit.
Is the concurrency limit the same as average concurrency? No. The limit is the cap the system enforces; the average is what agents really handle, and it is usually lower because chats start and end at different times.
Does higher concurrency hurt customer satisfaction? It can, because each customer waits longer between replies while the agent answers others. Track CSAT and reply gaps alongside throughput when you raise the limit.
Can an agent go over the limit? Yes, in most tools. Limits stop automatic assignment, but manual assignment, and in some tools reopened conversations, can push an agent above it.
How does concurrency affect how many agents I need? Higher average concurrency means fewer agents for the same volume, but the gain is cut by longer handle times. Use average concurrency, not the limit, in staffing calculations.
Related terms
- Erlang C — the staffing formula that needs a concurrency adjustment for chat.
- Average handle time — the number that rises as concurrency rises.
- Chat queue — where chats wait when every agent is at the limit.
- Chat routing — decides which agent with spare capacity gets the chat.
- Live chat — the channel where concurrency applies.
- Service level agreement (SLA) — the response promise the limit has to support.
- Chatbot first response time — the wait customers feel when agents are full.
Sources
- Zendesk, Creating capacity rules to balance agent workloads — support.zendesk.com/hc/en-us/articles/4776409839770, read 17 November 2026. Source of the messaging capacity maximum of 20, the separate count for messaging and live chat, the default rule, one rule per agent, the inactive-conversation setting, manual assignment exceeding capacity, and plan availability.
- Intercom, Balanced assignment deep dive — intercom.com/help/en/articles/6553774-balanced-assignment-deep-dive, read 17 November 2026. Source of workspace, teammate and inbox limits, workspace-wide counting, conversations waiting in the team inbox, manual assignment bypassing the limit, and the Expert plan.
- Intercom, Inbox assignment limits — intercom.com/help/en/articles/12960865-inbox-assignment-limits, read 17 November 2026. Source of how inbox limits combine with the teammate limit.
- HubSpot, Configure chat capacity limits for agents in your account — knowledge.hubspot.com/inbox/configure-chat-capacity-limits-for-agents-in-your-account, read 17 November 2026. Source of the default of 3, the maximum of 99, Service Hub Enterprise, unassigned chats at capacity and the visitor-facing max capacity options.
- LiveChat, Queue configuration — livechat.com/help/queue-configuration/, read 17 November 2026. Source of the per-agent chat limit and visitors entering the queue when all accepting agents are at their limit.
- Call Centre Helper, Erlang C for Chat? (Jonty Pearce) — callcentrehelper.com/erlang-c-for-chat-206984.htm, read 17 November 2026. Source of the concurrency factors, the quote on handle time rising with concurrency and the 3-or-4-chats example.
- Chatbotscape evaluation methodology. /methodology (continuously updated).