Chatbots are often sold as complete replacements for service desks. In practice, they are best treated as structured front doors: good at intent capture, routine deflection, and routing — weak at nuance, exception handling, and trust-heavy conversations.
Useful expectations: clear handoff to humans, transparent limits, measurable deflection quality, and alignment to ITIL/ESM workflows. Poor expectations: “zero agents,” unsupervised answers on policy edge-cases, and vanity automation metrics.
Related reading: Writing on ITSM and customer care, and The Future with Chatbots.