Call centres with AI agents on the line
Customer service and call centres appear on 25 recent papers, and 10 of them carry a chatbot, automation or language-model keyword. Firms now route calls through voice agents that handle part of each conversation and hand the rest to people. How should queues, staffing and handovers be designed when the first responder is a model?
What prompted this
Since October 2025, 25 papers carry a call-centre, contact-centre or customer-service keyword (11 in management journals, 4 in operations research): "customer service" on 12, "customer service automation" on 2, "customer service employees" on 2, "call centers" on 2. Ten of them also carry a queue, staffing, chatbot, AI, automation, language-model or routing keyword. The queueing models that once dominated call-centre research have not yet absorbed AI agents.
Questions
- When an AI agent resolves a share of calls and escalates the rest, how do waiting times and staffing needs change, and does the escalated traffic become harder per call?
- What is the optimal point to hand over to a person, and does the answer differ for customers who want to leave?
- Do AI-first systems lower costs more than they lower satisfaction and retention?
Data that could help
Call logs with routing, handover and resolution data from contact centres that deployed AI agents, and customer retention records.
Next step
Fit a queueing model with an AI first stage to call logs from one contact centre and test its staffing predictions against observed waiting times.