Why Tiers Break
Tier labels exist because companies that share a stage and business model tend to share an evaluation pattern. AI-native companies screen for whether you can design around model uncertainty. Growth-stage platforms screen for whether you can build a function while shipping product. Enterprise platforms screen for coherence across surfaces at scale. Regulated companies screen for whether you understand the cost of a confident wrong answer reaching a user.
Knowing the pattern saves you from walking in cold. But a single company can run multiple roles with unrelated evaluation frames. OpenAI's Growth design lead gets evaluated on adoption funnels and retention loops. Its Identity role gets evaluated on permissions architecture and agent-agent trust. Same tier label, completely different proof requirements.
The tier playbooks that follow give you the default evaluation frame for each archetype. Use them to orient. Then check the specific role: what does it own, and who outside Design gets hurt if it fails? Prepare for that answer.
When Roles Break Their Tier Pattern
The tier playbooks give you a default evaluation pattern for each company archetype, and most roles follow it. The ones that don't will screen you out before you understand why: you show up with enterprise-scale proof for a hiring manager who needed AI-product fluency, or you lead with velocity for a panel probing your judgment about consequences. This piece classifies four deviation categories drawn from current postings on your target list — AI-native roles inside enterprise structure, growth-stage roles with regulated consequences, enterprise roles with startup mandates, and regulated roles at shipping speed — with the recognition cues you can read off the posting and the positioning shift to make before the first conversation.
The Mandate Checksum — Three Readings That Confirm or Break a Tier Label
You've labeled the company AI-native, growth-stage, enterprise, or regulated, and that label is now controlling which version of your background you lead with. If it's wrong, you'll spend thirty minutes selling the solution to a problem the company doesn't have. Three readings, extracted through ordinary questions during a recruiter screen or hiring manager call, will confirm or break the label while you're still on the phone. One of them carries more weight than the other two. This piece covers detection only: what to ask, what to listen for, how to score what you hear.
Evidence Pivot Protocols
You prepared for the tier. The interviewer is evaluating something else — not one off-pattern question, but sustained attention to a priority your preparation didn't anticipate. Six reference cards for the most common deviations: governance questions inside AI-native conversations, speed questions inside enterprise ones, title-level mismatches, craft-intensity overrides. Each card gives you the trigger phrases to recognize, the evidence to substitute, the transitional language that preserves continuity, and the confirming question that tests whether you read the shift correctly. Three moves, under sixty seconds.
Deviation Map — Eight Roles That Break Their Tier Pattern
Tiers predict what an interview loop scores for. Eight roles on your active list break the prediction. OpenAI Growth screens for adoption economics, not model-behavior judgment. Suno wants a creative-product leader who builds with AI as a medium, not someone who designs how users supervise it. Capital One's Director mandate is organizational transformation — the role doesn't own a customer-facing AI product at all. Each deviation means your default evidence order is wrong for the room you're walking into. This piece groups all eight under four deviation patterns, with lead evidence, checksum questions, and the positioning mistake each deviation invites.