Concurrent calls (concurrency)
Concurrent calls are phone calls active at the same time, and concurrency is the number of those overlapping calls in progress at once. A system needs enough capacity to support its peak concurrency without delaying or rejecting new calls.
Concurrency is a peak-capacity measure, not a total-volume measure. A business may handle many calls over a day while needing little concurrent capacity if those calls are spread evenly. The same daily volume can require much more capacity when callers arrive together, such as after an advertisement, during a service outage, or at the start of an outbound campaign.
Each active call uses telephony, audio processing, and application resources until it ends or transfers away. When demand reaches an account, carrier, or system limit, another call may be queued, sent to voicemail, forwarded elsewhere, or rejected, depending on the configured call flow. Limits can exist at several points, including the phone number, carrier connection, SIP trunk, AI platform, and downstream integration.
For AI phone agents, the key planning input is the busiest expected interval rather than the monthly call count alone. Teams should consider arrival patterns, average handle time, seasonal spikes, retry behavior, transfers, and how long callers remain on hold. Outbound campaigns need particular care because a rapid dialing schedule can create a larger overlap than the underlying contact list suggests.
Concurrency also differs from calls per second. Calls per second describes how quickly new calls begin, while concurrent calls describes how many remain active together. A burst of short calls may create a high start rate with modest concurrency. Longer calls can create high concurrency even when new calls arrive slowly.
Capacity planning should include failure behavior. If the primary path is full, decide whether callers should hear a busy signal, enter a queue, reach a fallback number, leave voicemail, or try again later. Those choices affect caller experience and may change both handle time and the next traffic peak.
When comparing platforms or telephony setups, ask where concurrency is enforced, whether transferred or held calls continue to count, and how limits change. Measure actual peak overlap after launch and revisit the estimate when call duration or traffic sources change. Average usage can look comfortable while a short peak still overwhelms the available call paths.