The tier playbooks give you a default evaluation pattern for each company archetype, and the default holds most of the time. When it doesn't, the cost is concrete. You prepare enterprise-scale proof for a hiring manager who needed AI-product fluency. You lead with shipping speed for a panel that was testing your judgment about consequences. Nobody tells you which frame you were being read against.
Issues #7 and #9 laid out the standard sequences. This one covers roles that borrow their evaluation logic from a tier other than the one their company belongs to.
Detecting a deviation before it burns a cycle
Start from the company's tier label. The posting language tells you whether this particular role departs from it. The first conversation confirms.
Three things should describe the same job: what the posting says you own, what the interview process will actually test, and what one real decision in this seat would require you to make. When those three point at different tiers, the role is deviating from its company's pattern.
Issue #8 covered how to surface real decision rights once you're on the phone. This piece gives you reason to suspect the mismatch before you get there.
AI-native role inside enterprise structure
Hiring posture. Established enterprise platform. This one seat owns a product surface where AI is the interaction model rather than a feature. The company hired for scale and coherence everywhere else in its design org. Here it needs someone who can design human-agent experiences that don't exist yet.
How to spot it. Read the posting against the company's own standard design job language. Multimodal interaction, human-agent workflows, technical prototyping, code contribution — any of these inside an enterprise design posting is a deviation signal. Atlassian's Loom Lead Product Designer posting asks for multimodal human-agent experiences and occasional direct fixes to pull requests, none of which appears in Atlassian's standard design interview handbook. Capital One's AI in Experience Design director role names Claude, Gemini, Claude Code, and Windsurf as expected tools. That level of specificity tells you the evaluation will test whether you've built with these, not whether you have a point of view about them.
The timing pattern. These roles tend to surface four to twelve weeks after a company announces a move from isolated AI features toward an agent or workflow platform. Atlassian announced Loom workflows converting recordings into Jira and Confluence actions in March 2026; the human-agent design role appeared in June. Capital One published Context Specs for reusable agent skills in June; the design director role appeared in August. Two sequences is not something you can plan a search around. Render's agent-experience design role appeared before its most visible product announcement rather than after. Check the careers page when the product news lands; don't build a timeline on it.
How the evaluation borrows from both tiers. The interview architecture stays enterprise: portfolio review, cross-functional panel, values interview, calibration session. The content inside it shifts. Issue #9's split between AI-tool fluency and AI-product fluency decides which version you're facing. Tool fluency is whether you work with these systems daily; product fluency is whether you can design the behavior of one. Capital One's role tests the first, plus organizational transformation — can you change how 60,000 people work? Loom's tests the second. Same deviation category, different test.
One gap I can't close from public evidence: no enterprise company on your list has published a modified interview format for these roles. Whether you'll face a prototype exercise, a code-adjacent artifact review, or a live AI-product case is unknown. Ask the recruiter during scheduling: "Will any stage assess a working AI prototype or code-adjacent artifact, or will the AI-specific material come up through the standard portfolio and competency interviews?" The answer changes your prep entirely.
Positioning move. Lead with the Trust essay and Agentic Labs as AI-product proof, then bridge to Alibaba for enterprise-scale credibility. TinyFish gives you practitioner depth on agent traces, auditability, and governance. Use it in conversation, past tense. It isn't in your published portfolio, so a hiring manager can't verify it by looking. For Capital One's tool-fluency emphasis, the TinyFish production experience carries more weight than the Agentic Labs consumer apps. For Loom's product-fluency emphasis, Agentic Labs and the five handoffs framework lead.
A note on the forward-looking artifacts here. An enterprise hiring committee may not have the technical fluency to evaluate them the way an Anthropic panel would. If the panel includes AI/ML engineers or a product leader from the AI surface, lead with the artifacts. If it's staffed from the broader enterprise design org, the Trust essay's accessible framework and the shipped Agentic Labs apps are the safer lead. They prove AI-product fluency without asking the evaluator to judge frontier design thinking they may not be equipped to assess.
What gets you killed.
Leading with Alibaba scale as your headline. The enterprise panel will appreciate it. The hiring manager for this seat needs AI-product proof first. Alibaba is support.
Presenting AI as a strategic interest. "I'm deeply interested in AI" reads as "I haven't built anything." TinyFish production work and the live Agentic Labs apps close that gap in one sentence.
Ignoring the enterprise decision structure. The mandate is AI-native; the hiring committee is still enterprise. You need cross-functional influence proof alongside AI fluency, and Thermo Fisher's multi-stakeholder coordination supplies it.
Red flags for you. The hiring manager describes frontier AI work but the VP above them evaluates for platform coherence — the mandate is real to the team and not to the org. No one from AI/ML engineering appears in the interview loop, which suggests the role designs for AI without access to the people building it. The role reports into the enterprise design org rather than the AI product team, and nobody can name a recent product decision this seat would have owned that went against the enterprise default.
Growth-stage role with regulated consequences
Hiring posture. Growth-stage speed — ship, build the function, player-coach. But the product touches domains where errors cause real downstream harm: clinical decisions, financial obligations, legal exposure. The growth-stage playbook says lead with 0-to-1 builds and velocity. The consequences add a second filter: does this person understand decision gates, exception handling, and the cost of being wrong?
How to spot it. The posting mixes growth-stage verbs (build, ship, iterate, navigate ambiguity) with consequence nouns (clinical outcomes, provider workflows, patient safety, regulatory requirements). Headway announced on March 31 that the Tezi team had joined to advance human-centered AI in mental healthcare, then described internally developed healthcare-safe agents on April 13 (Headway blog). A cluster of provider-product design roles, including Provider CRM, appeared in early August. That's growth-stage velocity applied to workflows where a design error affects patient care. The four-month gap matches the timing pattern from the first category, but the evidence is thinner: the link is to Headway's blog index rather than individual announcements, and the connection between the AI announcements and the design roles is inferred from timing, not confirmed by anything Headway has said.
How the evaluation borrows. Pace and structure stay growth-stage — faster loops, fewer rounds, emphasis on portfolio speed. The portfolio interrogation is where the regulated dimension shows up. Expect questions about edge cases, about what you decided not to ship, about how you traded speed against safety. The transition map from Issue #7 — cost of failure, detectability, reversibility — is the vocabulary this evaluation runs on, whether or not the interviewer names it.
Positioning move. Lead with Thermo Fisher (pharma, $20M margin, exception-first design) and Red Cross (mission-critical, six systems to one, national deployment). Both prove you shipped fast where consequences were real. Bridge to TinyFish for AI production currency. The Allē dual-surface redesign (30M members, 40K providers) maps directly onto companies with a provider-plus-end-user split like Headway's.
What gets you killed.
Speed metrics without consequence awareness. "We shipped in three months" needs a second half about the cases where speed would have caused harm. Without it, you read as reckless in this room.
Compliance caution as your lead. Fatal in the other direction. These are growth-stage companies that need someone who ships fast and knows where to slow down. Thermo Fisher's twelve months to 100% partner adoption demonstrates both at once.
Treating the regulated dimension as something you navigated around. Present compliance as design territory you worked in, not an obstacle you routed past.
Red flags for you. The regulated dimension is used to justify decision cycles that contradict the stated mandate — "we move fast, but compliance review takes six weeks" means your velocity is someone else's to grant. Design reports to product and product defers to compliance on anything consequential, which shrinks your scope to the surfaces where errors don't matter. The company describes the regulated work as transitional ("once we get through certification..."), which usually means the constraint is permanent and the role's scope is defined by it whether or not anyone says so.
Enterprise role with startup mandate
Hiring posture. Established platform, but this seat owns something that doesn't exist yet: a new product surface, a new business line, a net-new experience vision. The rest of the design org was hired for platform coherence — this role needs a builder.
How to spot it. The posting uses 0-to-1 language inside an enterprise context: build from scratch, define the vision, new product area, greenfield. Run the checksum and the deviation shows. The posting describes startup-level ownership, the assessment still includes enterprise-style cross-functional panels, and the first real decision requires navigating platform constraints, stakeholder alignment, and inherited technical debt.
This category is thinner on your current target list than the others. Most enterprise 0-to-1 mandates right now are AI-related and collapse into the first category. The pure version — an enterprise company building something genuinely new that isn't primarily an AI product — is uncommon in this market. When it appears, the positioning error is specific and expensive.
How the evaluation borrows. Structure stays enterprise. The portfolio review shifts weight toward what you built from nothing rather than how you improved something existing. The cross-functional panel is the variable that matters. Staffed from the new product area, it evaluates builder instincts and tolerance for ambiguity. Staffed from the existing platform — which happens often, because the new area may not have senior people yet — it evaluates coherence and organizational fit. Ask during scheduling who's on the panel and where they sit. That answer tells you which version of your story to prepare.
Positioning move. Lead with Red Cross (0-to-1 in six months, national deployment, regulatory compliance from day one) and Thermo Fisher (0-to-1 in twelve months, 100% partner adoption). Bridge to Alibaba to prove you can build from zero inside a large organization with existing constraints. That combination is the differentiator; most candidates at this level can show one side or the other.
What gets you killed.
Leading with platform coherence and systems thinking. That's what the company hired for in its other design roles. This seat wants a builder first.
Presenting as a pure startup operator. The enterprise decision structure requires proof of organizational navigation. If your 0-to-1 stories contain no stakeholder alignment, no inherited constraints, no cross-functional negotiation, the panel will doubt you can operate inside their organization.
Describing your vision without naming the constraints you'd inherit. The posting says greenfield. The company has platform dependencies, shared design systems, and alignment requirements that bound what "from scratch" means. The Alibaba mandate-building narrative — naming a structural gap, then building the case across three cross-functional sprints — shows you know how to make something new inside something old. Show only the vision and the panel hears someone who will fight the organization instead of building through it.
Red flags for you. No dedicated engineering team for the new product area, which puts your roadmap behind someone else's. A panel staffed entirely from the existing platform whose questions focus on how the new product will integrate rather than what it should become. A hiring manager who can't name a recent decision where the new area chose differently from the platform default and the choice held. In each case the startup mandate exists on paper and the organization hasn't made room for it.
Regulated role with growth-stage velocity
Hiring posture. The product carries regulatory, compliance, or trust consequences, and the company ships at growth-stage speed. The evaluation tests domain seriousness and building velocity together.
How to spot it. The posting combines compliance or trust language with product-building language: 0-to-1, ship nondeterministic AI products, complex workflows, high-stakes environments. The timing pattern mirrors the first category — these roles tend to appear after a company expands from a focused compliance or trust product toward a broader platform. Ambience went from clinical documentation to a full healthcare AI platform in April 2026; the Staff Product Designer role appeared in June. Vanta announced its Agentic Trust Platform in November 2025; the AI-Powered Workflows role appeared in February 2026.
Two subtypes that evaluate differently. Reading Stripe Risk against Vanta's design roles shows a material split.
Risk as platform infrastructure (Stripe): the designer creates coherence across a mature financial platform — onboarding, compliance, account health, new products, AI-driven workflows, shared standards. Regulation is a constraint the product operates within. The evaluation weights systems thinking, cross-team influence, and platform-level craft. Complex-systems experience matters more here than prior financial-risk experience.
Compliance workflow as product (Vanta, Ambience): compliance or trust management is the product. The designer invents workflows for practitioners managing evidence, controls, risks, and approvals. The evaluation weights customer discovery, workflow modeling, and — in Vanta's current posting — experience shipping nondeterministic AI products. Vanta's VP of Design has described the work as adapting to customers' existing workflows and pressure-testing concepts through qualitative conversations, which puts more weight on domain familiarity than the platform subtype does.
Both subtypes will test your judgment about when to move fast and when to slow down. The positioning lead is what differs.
Positioning move. Platform subtype (Stripe): lead with Alibaba for platform coherence at scale and Thermo Fisher for the multi-stakeholder regulated ecosystem, with TinyFish adding AI production currency. Workflow subtype (Vanta, Ambience): lead with Thermo Fisher and Red Cross as mission-critical 0-to-1 builds in regulated domains, then the Trust essay and Agentic Labs for AI-product fluency. The AI evidence stack from Issue #9 applies. TinyFish is something you say in the room; it isn't something they can look up.
What gets you killed.
Treating both subtypes identically. Stripe probes for platform-scale systems thinking. Vanta probes for workflow invention and AI-product judgment. Wrong emphasis, wasted evidence.
Presenting compliance expertise as your primary credential. Both subtypes want builders who understand consequences. Your 0-to-1 builds in regulated domains are the proof.
Ignoring the AI dimension. Stripe and Vanta both mention AI-driven or nondeterministic product work explicitly. If your portfolio presentation says nothing about designing for AI behavior, you've missed a filter that exists even though the title doesn't say AI.
Red flags for you. In the platform subtype: design decisions require policy sign-off and the policy team runs on a different clock, which means the velocity in the posting may not survive the compliance review cycle. In the workflow subtype: internal growth-stage speed collides with compliance customers who expect enterprise-grade stability, and the tension lands on the designer. In both: the role's scope is set by what policy allows rather than what the product needs. Ask who resolves disagreements between product direction and regulatory interpretation. If the answer is "legal" or "policy" with no qualification, design influence has a ceiling the posting didn't mention.
When deviations stack
A role can show more than one deviation. Stripe Risk is an enterprise role with a startup mandate (a new product pillar inside a mature platform) and a regulated role with growth-stage velocity (financial compliance consequences at shipping speed). Identify the primary one — whichever most changes the evaluation from the company's default — and position for that. The secondary becomes a supporting proof point.
To pick: ask which mismatch would hurt you most if you prepared for the wrong one. For Stripe Risk, that's the regulated-domain judgment. Stripe already knows you can operate in a large organization if you've cleared their standard screens. What they need to confirm is whether you can make high-consequence product decisions at speed.
What the taxonomy doesn't resolve
The public evidence is good on what these roles advertise and thin on how they actually evaluate. No enterprise company on your list has published a modified interview format for a deviation role. The checksum gives you a detection mechanism; confirming a deviation still takes a conversation. Two questions do most of the work:
- "What does the interview process look like for this specific role?" — surfaces any modified assessment stages.
- "Can you walk me through a recent product decision this role would have owned?" — surfaces what the seat is actually accountable for, which may not match the posting.
Both connect back to the authority-reading framework in Issue #8. The taxonomy tells you when to suspect a mismatch. The first conversation tells you whether you were right.
Quick-Take Cards
Card 1: AI-Native Role Inside Enterprise Structure
Hiring posture: Enterprise company, but this seat owns a human-agent product surface. They're buying AI-product fluency inside an enterprise decision structure. Lead pillar: Trust essay plus Agentic Labs for product fluency; Alibaba as enterprise-scale support; TinyFish as spoken production credibility. Top landmine: Leading with Alibaba scale. The hiring manager needs AI proof first, scale proof second. Opening question: "Will any stage assess a working AI prototype or code-adjacent artifact?" Recognition cue: The posting names specific AI tools, multimodal interaction, human-agent workflows, or code contribution at a company whose other design roles mention none of these.
Card 2: Growth-Stage Role With Regulated Consequences
Hiring posture: Growth-stage speed in a domain where design errors cause real harm. They're buying someone who ships fast and knows where to slow down. Lead pillar: Thermo Fisher (pharma, exception-first) and Red Cross (mission-critical, national). Allē for provider/end-user relevance. Top landmine: Speed metrics with no consequence awareness. "We shipped in three months" needs "and here's what we decided not to ship." Opening question: "How does the team decide when a feature needs additional review before shipping?" Recognition cue: Build/ship/iterate language sitting next to clinical outcomes, provider workflows, or patient safety.
Card 3: Enterprise Role With Startup Mandate
Hiring posture: Established platform, but this seat owns something that doesn't exist yet. They're buying a builder who can navigate enterprise constraints. Lead pillar: Red Cross and Thermo Fisher for 0-to-1; Alibaba for building inside a large organization. Top landmine: Leading with platform coherence. This role wants a builder first, an enterprise operator second. Opening question: "Is this a new team, or an existing group with a new mandate?" Recognition cue: Build from scratch, greenfield, new product area, define the vision — at a company whose other design roles emphasize platform and systems.
Card 4: Regulated Role With Growth-Stage Velocity
Hiring posture: Consequential domain at startup speed. Two subtypes: risk as platform (Stripe), compliance workflow as product (Vanta, Ambience). Lead pillar: Platform subtype — Alibaba plus Thermo Fisher. Workflow subtype — Thermo Fisher, Red Cross, Trust essay. Top landmine: Treating both subtypes the same. Stripe probes platform systems thinking; Vanta probes workflow invention and AI-product judgment. Opening question: "Is the regulated dimension a constraint the product works within, or is it the product?" Recognition cue: Compliance or trust language beside 0-to-1 building language, with a mention of AI-driven or nondeterministic work.
- Render posted before announcing: Render's agent-experience design role appeared days before its most visible product announcement, making the role itself a leading indicator of the company's agent strategy rather than a trailing one.
- Vanta's design philosophy surfaced: Vanta's VP of Design described the team's approach as adapting to customers' existing workflows and pressure-testing concepts through qualitative conversations, which tells you what the portfolio interrogation will probe before you walk in.
- Atlassian's interview handbook still: Atlassian's published design interview process has not been updated to reflect AI-specific assessment stages, even as its Loom posting asks for multimodal human-agent experience and code contribution — confirming the gap between role requirements and documented evaluation format.
- Capital One's AI fluency push: Capital One reported making an enterprise AI learning hub available to more than 60,000 associates before posting the AI-in-Experience-Design director role, which means the organizational transformation mandate in that posting is already underway, not hypothetical.