Concurrency is the number of calls a system can hold at the same moment. In outbound dialing, one concurrent line is equivalent in capacity to one agent seat. Concurrency limits appear at every layer: the dialer, the SIP trunk, the carrier and, for voice bots, the number of bots running on the campaign.
When a floor says it is running 50 lines, it usually means 50 calls can be in conversation at once, not 50 dials.
Concurrency on a live campaign
A floor in Noida runs 30 fronter seats. With predictive dialing, the dialer may have far more than 30 calls ringing at once, but only 30 can be in conversation. If the SIP trunk only allows a set number of channels, the dialer cannot place more calls than that, however high the dial level is set. Every layer has to be sized for the peak.
For voice bots, concurrency is usually per bot: one bot holds one conversation at a time, the same as one human seat. Adding capacity for a surge, AEP for example, means adding bots or seats, not just raising the dial level.
Common mistakes
Sizing trunk channels for the number of agents instead of the number of simultaneous dials.
Raising the dial level when the real limit is concurrency somewhere else in the chain.
Forgetting the closer side. Fronter concurrency is wasted if the closer queue can only take a few transfers at once.
How it shows up in reports
Real-time dialer screens show calls in progress against agents or bots available. If calls in progress stay below the number of seats while agents wait, the dial level is too low or the trunk is at its limit. If transfers queue on the closer side, the closer team is the constraint. Peak concurrency per hour, plotted across a week, tells you how much capacity a campaign actually needs.