AI is becoming common in software teams, but using it safely in customer support is harder than simply connecting a chatbot to documentation. In this lightning talk, I will share what I learned while building an agentic support system for WordPress plugin support.
The system was designed to help support engineers investigate customer issues by searching verified documentation, checking internal notes, analyzing diagnostic reports, separating confirmed facts from assumptions, and drafting accurate replies. The main lesson was simple: AI can be useful in support only when it is forced to verify before answering.
I will cover the key decisions behind the workflow: how to structure a knowledge base, why source priority matters, how to reduce hallucinated answers, when to escalate instead of guessing, and where human judgment still matters most.
This talk is for plugin teams, agencies, support engineers, and developers who want to use AI in WordPress workflows without losing accuracy or trust. Attendees will leave with a practical mental model for building AI-assisted support systems: verify first, cite sources, separate assumptions, and keep humans in control.
Did you attend this session? Leave feedback.