When a Longest-Idle style routing algorithm selects an agent for a new digital interaction (chat/email) and multiple agents tie on every other ranking factor (skill profile, rank, role), the routing engine falls back to comparing how long each agent has been idle since their last interaction ended.

The issue is that idle time appears to be tracked using a single timestamp that only updates at the start or end of an interaction — and for voice calls, this timestamp seems to update only when the call ends, not when it begins. Digital interactions, by contrast, appear to update the timestamp both when they start and when they finish.

This creates an inconsistency: an agent who is currently on an active voice call still shows an old "last activity" timestamp from before the call began, making them look like they've been idle longer than they actually have. Meanwhile, an agent who just finished a digital interaction is marked as recently active. As a result, the routing engine can select the agent who is busy on a voice call over an agent who is genuinely free, simply because the busy agent's stale timestamp makes them appear to be the "longest idle" option.

Requested enhancement — either of the following would address this:

Update the idle/recency timestamp for voice interactions when the interaction begins (not only when it ends), so it reflects real-time availability the same way digital interactions already do; or
Provide an admin-configurable setting, at the account or queue level, for how "idle" is measured for routing purposes — for example: recency across all channels combined, channel-specific recency, or excluding any agent currently engaged in any interaction from being considered "idle" at all, regardless of timestamp.

This would prevent new digital interactions from being routed to agents who are actively on a call while other agents are genuinely available, improving routing accuracy and customer response times in queues where concurrent voice and digital interaction handling is enabled.