One question filters every company on your list: does design own a consequential decision surface, or does it decorate one?
The test changes by tier. AI-native: does design own decision gates — what the model does, not how the output looks. Growth-stage: does design own exception paths — what happens when the system fails or the user disagrees. Enterprise: does design arbitrate coherence — who wins when business units conflict on experience. Regulated: does design own automation boundaries — where the machine stops and the human decides.
Pull this up before any engagement. Use it in order.
Phase 1: Public Signals
Ten minutes. LinkedIn, job postings, leadership pages, company blogs. You are reading for ownership evidence, not design-culture enthusiasm.
Three categories. Green means ownership evidence exists. Yellow means the role likely packages someone else's decisions. Check is the specific thing to verify before you spend time on a conversation.
AI-Native
- Green: Posting or company content says design affects model behavior, evaluation criteria, autonomy thresholds, or recovery paths. OpenAI's Model Designer posting is the clearest current example — the designer collaborates with researchers to "design model behavior." That is system-level authority.
- Yellow: Design language stays at "cohesive voice," "delight," or "user experience" while Product, Research, or Safety owns behavior decisions. The role packages outputs.
- Check: Whether the company's governance framework names design as a participant in behavioral decisions. If design is absent from the stakeholder list, that is your answer.
Growth-Stage Platform
- Green: Design reports to a CPO who reports to the CEO. Posting names cross-pod coherence, exception-path ownership, or human-agent workflow rules. Amplitude's structure is the current example: CPO oversees product, design, and growth, reports to CEO. The design leader posting names coherence across autonomous pods.
- Yellow: Design reports into product with no design executive. Role is measured by pod throughput, not by exception-path ownership or cross-surface coherence.
- Check: Whether the company's product content describes AI as a faster chart — screen-level — or as a product loop with autonomy thresholds and learning — system-level. The gap between marketing language and product-blog language is the most reliable tell at this tier.
Enterprise Platform
- Green: A design-system or platform-design leader owns adoption, governance, accessibility, AI-readable patterns, and cross-product coherence. Atlassian's design system is treated like a product with a roadmap, company-level goals, and measurable AI performance gains (52% accuracy improvement in AI calls after structuring foundations for AI consumption). That is infrastructure authority, not component documentation.
- Yellow: Design absent from the executive leadership page. Role language emphasizes "influence" and "storytelling" but never names arbitration rights when business-unit roadmaps conflict with system coherence. Enterprise platforms routinely give design scope without sovereignty. The pattern is consistent: verified public leadership pages for Salesforce, Adobe, Intuit, and Microsoft AI do not show design executives at the top layer, even when those companies fund large design functions inside product or BU structures.
- Check: Whether design-system language stops at visual consistency and reusable components, or extends to AI-readable semantics, code connection, adoption metrics, and cross-product arbitration.
Regulated (Healthcare, Compliance, Vertical SaaS)
- Green: Role reports to CPO or has direct CEO access. Owns product design plus research or design engineering. Names clinical or compliance outcomes. Maven's VP of Design reports to the CPO with a direct line to the CEO and owns product design, brand design, and user research.
- Yellow: Executive title, but automation boundaries, clinical escalation, launch gates, or auditability are owned entirely by Product, Clinical, Legal, or Engineering.
- Check: Whether published content describes design decisions affecting workflows and control surfaces, or only the member-facing UI layer. Vanta's Trustcraft post is the benchmark: traceability, eval harnesses, rollback safety nets, explicit rejection of "AI sprinkle." Gusto's AI principles name intelligent escalation and preserved agency as design questions. Companies that frame these as engineering-only problems are telling you exactly where design stops.
Cross-Tier Signals
Read the posting for negation language. When a company says what the role is not, they are telling you what went wrong last time. Amplitude's posting describes the role in opposition to an executive who manages up and delegates down, to design ops, and to an approval committee. That is the clearest brief you will get from a public document. Mirror the negation back in your positioning.
Watch for a title change between the departing and incoming leader. VP of Design becoming Head of Product Design under a CPO who owns PM, design, and growth is a scope and political repositioning, not a title preference. Reconstruct what changed before you accept the new frame.
One speculative signal worth monitoring: when a current design leader's LinkedIn shifts from full-time to "advisor," or they start posting thought-leadership content after months of silence, a departure is likely in motion and the mandate is about to be redefined. I cannot confirm departures from profile changes alone, but the pattern has preceded confirmed openings often enough to act on. Act by preparing, not by reaching out.
Phase 2: First Conversation
These are diagnostic probes, not interview questions. Ask them early. Listen for one thing: specificity versus abstraction. A real example with names and tradeoffs is green. A pivot to "we're building toward that" is yellow.
AI-Native
"When the model produces an output the user shouldn't trust, who decides how that gets handled?"
If the answer is Research or Safety with no design involvement, design decorates the trust surface. It does not own it.
"Can you walk me through a recent decision where design changed model behavior, not the interface around it?"
You want a named decision with a named tradeoff. "We're deeply involved in that process" without an example means they are not.
Growth-Stage Platform
"When two pods ship conflicting experiences, who arbitrates?"
If the answer is "the PM lead" or "we haven't had that problem yet," design does not own coherence. It requests it.
"What happens when a user overrides what the system recommends?"
This tests exception-path ownership. A designed recovery flow with escalation logic is green. "We log it" is yellow.
Enterprise Platform
"When a business unit's roadmap conflicts with platform coherence, who has the final call?"
Arbitration means the design leader can block a BU ship decision on coherence grounds. Advising means they write a memo. You need to hear which one.
"How does the design system team's authority work when a product team wants to deviate?"
Look for governance language: review, approval, exception process. "We try to influence" means the system is a suggestion.
Regulated
"Where is the line between what the system decides automatically and what requires a human?"
If the answer is precise and names specific workflows, design helped draw that line. If it defers to Legal or Clinical, design implements boundaries someone else defined.
"When a compliance or clinical requirement conflicts with the user experience you'd want to ship, how does that get resolved?"
You want to hear that design is in the room for that resolution. Not informed of the outcome afterward.
Pay attention to whether the hiring manager asks you "What would you have done differently?" versus "How would you have managed the team through it?" The first evaluates you as a product peer. The second evaluates you as a service provider. This predicts mandate scope better than title or comp.
Phase 3: Follow-Up Confirmation
You have public signals and first-conversation data. Now confirm or kill.
Confirm Across All Tiers
- Decision rights are funded. Budget, headcount authority, or explicit veto power over a defined surface. A title without budget is a costume. Ask what you would control.
- Reporting line matches authority. One layer from the CEO or CPO is green. Two layers is survivable if the middle layer is a genuine sponsor. Three layers is decorative regardless of title or comp.
- Predecessor departure is readable. If you cannot reconstruct why the last person left or why the role was created, you are accepting a mandate you cannot verify. Do that work before you proceed.
Kill Signals
- The hiring manager cannot name a specific business consequence that design decisions caused in the last six months. Not "design is important to us." A consequence. Revenue, retention, clinical outcome, compliance event. No consequence means no authority. (High confidence.)
- Design is absent from the interview loop for its own leadership role. If the panel is all product and engineering, design is a service function regardless of what the posting says. (High confidence.)
- Glassdoor or Blind reviews show repeated themes — 10+ reviews over 12+ months — of unclear ownership, feature-factory behavior, or frequently changing reporting structures. One review is noise. A pattern is signal. Vanta's recent reviews surface ownership and alignment concerns worth probing, though not yet at the density that confirms a decorative mandate. (Moderate confidence.)
- Public content describes design as "central" or "core" but the job posting uses product-manager language: OKRs owned by PM, design "supporting" product goals, success measured by feature velocity. The gap between the marketing and the posting is the real org chart. (Moderate confidence.)
Walk-Away Triggers by Tier
AI-Native. Walk away when design cannot affect model behavior, only the interface around it. Walk away when the governance framework lists every function except design as a behavioral stakeholder. (High confidence.)
Growth-Stage. Walk away when design reports into product with no design executive and the role is scoped to pod execution without cross-pod authority. Walk away when "coherence" appears nowhere — not in the posting, not in the conversation, not in the org structure. (High confidence.) Walk away when the CPO has never hired a design leader before and cannot articulate what decision rights the role will have. (Moderate confidence. This could also mean the mandate is genuinely open and shapeable. Probe harder before killing it.)
Enterprise. Walk away when the design-system team has no governance authority and operates as a service desk for product teams. Walk away when design leadership is absent from the executive page and no one in the interview process can explain why. (Moderate confidence. Some enterprise companies fund strong design functions inside BU structures without executive-page visibility. The burden of proof is on them to show you.)
Regulated. Walk away when automation boundaries, clinical escalation paths, or compliance launch gates are owned entirely outside design and the company sees no reason to change that. Walk away when the posting names "trust" but the first conversation reveals trust decisions are made by Legal, Clinical, or Engineering and handed to design for implementation. (High confidence.)
What the Role Must Give You
The mandate is real. Now test whether the role is worth taking.
Function-building authority. Can you hire, fire, set the quality bar, and define the operating model? Ask directly: "What would I have the authority to change in the first six months?" Vague answers here are not modesty. They are the answer.
Scope expandability. A Head of Product Design at a company about to add a platform layer, an AI product line, or a new market has natural expansion ahead. The same title at a company with a stable product surface and no adjacent ambitions is a ceiling. Identify which one you are looking at before you invest.
Reporting line to decision-makers. One layer is green. Two is survivable. Three is decorative.
Equity worth the role. Public salary ranges tell you cash. They tell you nothing about equity. At the offer stage, a company that sells "equity upside" but cannot explain its own 409A valuation, liquidity path, refresh cadence, exercise window, or how the grant was benchmarked for your level is filling a seat, not making a strategic investment in the person who fills it. Refresh grants are not guaranteed. Ask whether design leadership participates in refresh cycles at rates comparable to engineering leadership. The answer tells you where design sits in the company's actual hierarchy, which is the one denominated in dollars.
Ask whether the company can name what changes if design leadership is excellent versus adequate. If they cannot articulate the difference, they do not need what you do. Move on.
- AI-native title splits matter: Anthropic's Design Engineer posting explicitly deprioritizes people management while Suno's Head of Product Design asks for formal org scaling, so treat "AI-native" as two distinct mandate shapes (builder-stage vs. org-stage) rather than one tier.
- Overreliance research complicates trust: Buçinca, Malaya, and Gajos found that cognitive forcing reduced overreliance but users rated the interventions less favorably, which means "more transparency" is not a universal design answer and your trust-calibration argument needs to be more specific than that.
- Equity compression at later stages: Sequoia and Carta report that VP equity dropped slightly after the labor-market cooldown while AI-leader equity rose 25% year over year, so the equity conversation should include whether the company benchmarks design leadership against AI-engineering leadership or against a general executive band.
- NIST monitoring as diagnostic language: NIST's 2026 report identifies human factors monitoring as a distinct category for deployed AI systems, which gives you a concrete framework to reference when probing whether a regulated company treats post-deployment human oversight as a design responsibility or an engineering one.