Drafting authority is not spending authority
You're in the support studio. What you're doing this lesson: write down what the AI may draft, what a named person must approve, and what a human must handle — before the tired Friday ticket decides for you. That written split is the escalation boundary. You do not need the earlier support lessons to start.
Everything your AI-assisted queue produces belongs on exactly one of three lists. Writing the lists down — with an owner and a date — is the escalation policy. Not writing them down means the policy gets decided ad hoc, per ticket, by whoever is most tired.
- AI may draft. Replies, case summaries, tags, suggested canned answers. Draft, never send: everything here still gets a human look before a customer sees it. This list is long, and it's where the measured speed gains live (a large study of support agents found about 15% average productivity from an AI assistant — assist, not replace).
- Named approver required. Every customer-facing reply until that canned answer has passed its past-ticket tests and the policy grants it. And one category that never climbs the ladder: every refund and every adjustment, always. The AI drafts the refund reply; a human with refund authority executes the refund. Those are different verbs.
- Human only. Escalations, compensation decisions, anything touching legal or a regulator. The AI doesn't draft here — a model-written first draft of a chargeback response or a regulator letter anchors the human who edits it, and this is precisely where that first-draft pull costs real money.
Why refunds never graduate
The temptation is obvious: the refund-status canned answer just passed its tests, refunds under $20 are basically noise, why not let it close the loop? Because the two authorities are different in kind, not degree.
Drafting authority is about words being correct — checkable against a help article, lintable against a banned list, testable against past tickets. You can earn it with evidence.
Spending authority is about moving the company's money and making exceptions to its policy. Your company doesn't grant that by track record of good sentences — your team lead has a refund limit, and she writes well. A well-tested canned answer has exactly as much spending authority as a really good template: none. The standing rule: AI drafts, a human with authority executes. No test score converts one authority into the other.
This maps onto the legal reality too. The Air Canada chatbot case (Moffatt) was about a statement — and the company paid for it. Now imagine the bot doesn't just misstate the refund policy but executes the refund it invented. The blast radius goes from one tribunal claim to a ledger.
The boundary is a routing decision, made per ticket
The three lists are static policy. The live question, forty times an hour, is: which list does this ticket fall on? That's escalation — and here the market teaches it wrong. The helpdesk courses on offer treat escalation as a routing feature: pick the tier-2 queue from a dropdown. What it actually is, on an AI-assisted team, is a decision discipline: a small set of conditions, written so precisely that a new hire on day one reaches the same answer as your most experienced agent.
Klarna's failure territories tell you what the conditions must catch: emotional, multi-step, high-value. Next step writes those conditions down.