Advisory Panel
Each advisor works from a named perspective and answers in its own voice. Nobody argues against anyone else and no winner is declared — a chair reads the advice and writes up where it agrees and where it doesn't.
The Economistunit economics, second-order effects
The payback question is less interesting than the substitution question. If deflection works you don't save a headcount, you redeploy it, and the value shows up wherever that person goes next. Budget it as reallocation rather than savings, or you'll be measuring the wrong thing by Q3.
The Ethicistobligations, affected parties
Disclosure is the load-bearing question. A customer who doesn't know they're talking to a system can't calibrate how much to trust the answer, and the people most likely to be harmed by that are the ones least able to escalate. Whatever you deploy, the handoff has to be one action rather than a maze.
The Operatorhas actually shipped this
Everyone underestimates the content job. Someone owns the knowledge base or the whole thing decays within ten weeks, and it can't be the support lead in their spare time. If you can't name that person today, you're not ready to buy the tool.
The Skepticassumes the plan fails
Assume it goes badly. What's the tell, and how fast do you see it? If your only signal is quarterly satisfaction data you'll be six weeks into a trust problem before it's visible. Instrument escalation rate weekly and set the number that triggers rollback before you sign anything.
ChairSynthesis
Three of the four advisors land on a precondition rather than a decision: name the knowledge-base owner. The economist and the operator get there from cost, the skeptic from failure detection. The ethicist's point about disclosure is unrelated to the other three and nobody else addresses it, so it's better handled as a separate piece of work.
3 of 4 share a precondition1 unrelated concern0 direct contradictions