ai-assisted-resolution-strategy
/ai-assisted-resolution-strategy
Naming note. This skill was previously scoped as
ai-deflection-strategy; that name is retired and no alias resolves it — older plans and context records that name it refer to this skill. "Deflection" framing optimizes for the wrong metric (volume reduction) and is hostile to customers; "assisted resolution" correctly positions AI as a first-contact tool that escalates honestly when it hits its limits. The whole point of this skill is to enforce that honest-escalation posture structurally, not as a reminder.
Purpose
/ai-assisted-resolution-strategy produces a first-contact AI resolution strategy that treats the AI as the junior team member of a support org, not as a cost-reduction device. It bakes in a hard CSAT floor per segment, a three-trigger escape hatch that immediately hands off to a human, a never-deflect list of categories the AI may not attempt at all, and a mandatory rollback plan that fires automatically if the CSAT floor is breached.
The skill's one-sentence job: make sure the customer always has a clean, visible path to a human, and make sure the AI is measured against resolution quality rather than deflection volume.
When to Use
Invoke /ai-assisted-resolution-strategy when you need to:
- Launch AI-assisted support for the first time and you need the strategy document before picking a tool
- Trigger a quarterly CSAT review that has surfaced a degradation and you need to refresh the strategy (floors, escape hatches, rollback threshold)
- Add new languages to an existing AI resolution footprint (which reopens the multilingual detection rules for the escape hatch)
- Expand AI-assisted resolution into a new segment, especially a regulated or high-touch one where the previous floors were set for a different customer mix