Escalation routing works by monitoring conversations against a set of triggers configured in advance - sentiment scores dropping below a threshold, keywords signalling frustration, a query type outside the AI's scope, or an action that Dynamic Action Gating marks as requiring human sign-off. When a trigger fires, the system hands the conversation to a human agent along with the transcript, the AI's classification of the issue, and any data already collected, rather than simply dropping the customer into a queue with no context.
Good escalation routing matters intensely in MENA customer service because conversations often move fluidly between Arabic dialects, English, and mixed Arabizi, and a poorly briefed human agent can lose the thread a customer already built with the AI. In regulated sectors like banking or healthcare, escalation is also the point where compliance-sensitive decisions land in human hands. Preserving context at handoff keeps resolution times down and protects the customer experience across the WhatsApp and web chat channels Eshal typically deploys.
In Eshal: Agents receive the full transcript, the AI's issue classification, and any data already collected before the handoff, reducing agent handle time on escalated contacts by an average of 38 percent. Escalation triggers are configured per client based on sentiment, query type, and action-gating rules, so routing reflects each organization's own risk tolerance and service priorities rather than a generic default.