Day 1 · 09:00
Drag the sliders, switch the AI up a level, and watch the floor react — shorter calls, calmer agents, fewer callers left waiting. Every chart here is driven by the little cartoon team below.
Under the hood it's a tiny queueing model running at 30× speed: calls arrive at the rate you set, each agent can hold one call, and AI level changes three things — average handle time (transcription and suggested replies make calls shorter), wrap-up time (AI writes the call summary), and deflection (at higher levels, simple calls are resolved by self-service before reaching a human).
The escalation threshold is the trade-off dial: keep it low and the AI hands customers to humans quickly; push it high and the AI attempts more on its own — great for deflection, but past ~65% it starts failing calls that bounce back to the queue, and first-call resolution dips.
The autonomous agents are modelled on the ones in Microsoft Dynamics 365 Contact Center: the Customer Intent Agent studies past conversations to understand why customers call (smarter routing and article matches, fewer failed AI hand-offs), the Case Management Agent creates, updates and closes cases by itself (wrap-up time collapses), and the Knowledge Management Agent turns closed cases into fresh knowledge articles (resolution keeps improving). Each one you enable appears as a little robot doing its rounds — and nudges exactly the metrics it's responsible for.
Customer satisfaction tracks what callers actually feel: it climbs with instant AI answers and first-call fixes, and sinks with long queues, hang-ups and failed escalations.
Try this: load "Black Friday", leave AI on Basic, and watch the queue grow and both smileys wilt. Then switch to Full assistance and flip all three autonomous agents on — same team, completely different day. That gap is the whole argument for AI in customer service.