Your target list has fourteen qualifying roles right now. They share surface-level language — "AI," "Staff," "Product Design" — but they sort into five distinct accountability units, each buying different proof. Brex wants trust architecture. Ramp wants shipped product impact. Gusto has six openings across four of the five clusters. One version of your story sent to all six gets averaged to nothing. This scan breaks the posting language into firing conditions — what the hire gets blamed for if it fails — and maps each to the evidence from junochen.com that answers it.

The posting says "own," "define," "elevate," "transform." The executive who wrote it probably wants all of that. What they haven't done is transfer the budget, the decision rights, or the gates that would let you actually do it. This piece maps the gap between transformation language and transferred authority across three tiers — AI-native startup, growth-stage platform, enterprise — with a 60-second posting scan, conversation questions that test jurisdiction in real time, and walk-away rules for when the answer is clear.

The Capital One AI-in-XD Director search didn't start in May. It started in February, across three separate requisitions — and the language changes between them tell a consistent story: the committee moved away from executive presence and C-suite gravitas, toward implementation proof and systems-building. That shift took seven months and landed on a profile closer to yours than the original posting described. This piece reconstructs the full sequence, names what the committee likely learned, and tells you who to contact first.