## TL;DR
AI is the silent coach on every customer service call -- transcribing conversations in real-time, suggesting the best responses, detecting customer frustration, and scoring quality across every interaction. From Cresta to Gong, AI transforms call centers from cost centers into intelligence hubs.

## Core Explanation
Call center AI: (1) Real-time transcription -- ASR (Whisper, DeepSpeech) converts speech to text with <500ms latency. Speaker diarization tracks agent vs customer turns; (2) Agent assist -- NLP analyzes live transcript, surfaces relevant knowledge articles, suggests next-best-action, flags compliance violations, detects customer sentiment; (3) Post-call analytics -- auto-scores 100% of calls on quality rubric (greeting, empathy, resolution, compliance). ML identifies coaching opportunities and trending issues; (4) WFM -- ML forecasts contact volume per 15-min interval, auto-schedules agents with optimal skills mix.

## Detailed Analysis
Cresta: real-time agent assist using GPT-4. Features: instant knowledge surfacing (agent mentions product -> relevant docs appear), behavioral guidance ("slow down, the customer sounds confused"), and auto-generated call summaries. Cresta reports 15-25% AHT reduction. Gong: post-call revenue intelligence. Analyzes sales calls for deal risk (competitor mentions, pricing objections, lack of decision-maker involvement). Cogito: behavioral AI analyzing voice tone. Detects: agent burnout (flat tone), customer escalation risk (rising pitch), and conversational dynamics (interruption patterns). Speech emotion recognition: models trained on acted emotional speech (RAVDESS, IEMOCAP) classify 6-8 emotions. Challenges: real-world emotional expression is more subtle than acted datasets. WFM AI: combines historical call patterns + calendar events + marketing campaigns -> call volume forecast. Optimization: Erlang-C + ML for multi-skill scheduling.