You've done the failure diagnosis, built the evidence map, written the authority questions. You haven't contacted anyone.
This is the most common failure mode at your level. The analysis is real, and it's become a sophisticated reason to keep analyzing. The protocol below exists to break that cycle. It produces one output: contact, or the name of the specific gap preventing contact.
Fill four lines before you reach out to anyone at a target company. Each requires specifics drawn from diagnostic work you've already done. All four filled, you contact. If any line contains placeholder language or can't be completed, you name what's missing and either close that gap or move to the next target.
Line 1: What problem was this role created to solve?
What it measures. Whether you understand the company's fear — the specific thing that goes wrong if this seat stays empty another six months.
How to generate it. One sentence extracted from the failure diagnosis you already built for this company. Name the failure, not the aspiration.
Weak version: "Create a world-class provider experience."
Usable version: "Prospective patients fall out of fragmented intake workflows while providers spend time operating a digital front desk instead of delivering care."
The correction when you're stuck: ask what breaks, for whom, if nobody fills this role. If your answer could apply to thirty companies, it isn't specific enough. "Build trust in AI interactions" is a category that covers half your target list. "AI behavior in care delivery mishandles uncertainty and defers to human clinicians too late or not at all" describes one company's problem.
Line 2: Which piece of your portfolio evidence is most relevant, and why?
What it measures. Whether you can name one artifact and explain in a single sentence why its structure transfers to the problem in Line 1.
How to generate it. You already mapped your evidence to this company's tier and failure. Line 2 picks the lead and commits. The evidence layer that leads varies by tier: forward-looking design artifacts for AI-native companies, published case studies for enterprise and regulated companies, function-building proof for growth-stage companies. Your mapping work tells you which. Line 2 makes you say it out loud.
When your evidence map gives you two or three plausible leads, select for transfer precision over project prestige. Pick the artifact whose problem-shape matches Line 1's failure with the fewest qualifications in the "which maps to" clause. If you need two sentences to explain the transfer, the mapping is weaker than one that takes half a sentence. When two artifacts transfer equally well, pick the one the evaluator can inspect without you in the room — the published case over the verbal walkthrough, the interactive prototype over the static deck.
Weak version: "I'd lead with Alibaba."
Usable version: "Alibaba — redesigned an outdated B2B platform across multiple surfaces and stakeholder groups, which maps to creating a unified AI interaction direction across product teams with their own roadmaps."
The clause that starts with "which maps to" is what Line 2 exists to produce. If you can't write it, the evidence doesn't transfer as cleanly as you assumed. Pick a different lead.
One constraint that applies every time: the evidence you name must be something the evaluator can inspect. If it's not findable from your portfolio site, include the direct URL in your outreach. Reachable is not discoverable. A reviewer who can't find the artifact without a direct link can't evaluate it.
Line 3: What do you need to learn about whether design has real decision-making authority here?
What it measures. Whether you've identified the specific authority question the posting leaves unanswered and can ask it in a way that produces a diagnostic answer.
How to generate it. You already tested whether the role can deliver what it advertises. You read the posting for decision-rights language and found that it advertises influence without disclosing exercised authority. Line 3 converts that gap into a question about a concrete decision.
Weak version: "How design-led is the organization?" That question produces a sales pitch.
Usable version: "When the AI front desk could improve conversion but created a risk of an inappropriate patient interaction, who chose the automation boundary?" That question asks who decided, not how the company feels about design.
The mandate checksum established that a real decision story is more diagnostic than any amount of posting language. Line 3 is where you write the question that gets you that story.
Line 3 is a question you're bringing to the conversation, not an answer you need before it. You don't need to know the authority structure before outreach. You need to know what you're going to ask. If your question is specific enough that it could only be asked at this company about this role, Line 3 is complete. If it could be asked anywhere, it's not.
Line 4: What structural condition would cause you to downgrade this opportunity?
What it measures. Whether you've identified the specific structural arrangement that would make this role unable to deliver on its advertised mandate.
How to generate it. You already separated strategic presence from decision rights for this company. Line 4 names the trip wire: the structural condition you'd discover through Line 3's question that would change your assessment.
Weak version: "I'd prefer a team of at least five designers." That's a preference.
Usable version: "Design is accountable for the trust experience but cannot alter the model output, deferral rule, or release condition." That's a structural constraint — accountability without authority, which predicts whether a design leader will have impact or spend two years fighting for influence.
Your downgrade condition should describe a mismatch between what the role is accountable for and what it controls. If the role owns the outcome but can't change the inputs, that's structural. If you don't love the product category, that's preference. Preferences don't belong in Line 4.
The protocol against four live roles
Each demonstration shows the completed protocol against a live role, with the evidence layer that leads for that tier.
AI-native: Anthropic — Product Designer, Design Systems
The posting says the role covers "guidance, metadata, evals" for Claude-generated UI. That's a trust-architecture problem — defining the criteria by which an AI's outputs get judged — and it points the failure diagnosis toward drift and the absence of enforceable standards.
Line 1. Product teams and Claude's own generated UI diverge from the shared interaction system, and nobody owns the eval criteria that catch the drift.
Line 2. The Trust essay — its framework for evaluating AI-generated outputs against interaction standards maps directly to making a design system legible to Claude when Claude generates interfaces. The core problem is the same: the system must define criteria that constrain outputs it doesn't directly control. Forward-looking design-systems specifications support the systems-architecture dimension.
Line 3. When a Claude product team or Claude-generated interface departs from the system, who can require a correction before release, and does this role own the eval criteria used to make that call?
Line 4. Downgrade if the role can publish patterns and guidance but cannot change package behavior, Claude-facing metadata, adoption requirements, or release thresholds. That's a documentation role with a systems title.
Forward-looking design artifacts carry the heaviest differentiation load here. Production proof validates but won't differentiate you. Lead with the artifact that shows how you think about AI-legible design systems.
Growth-stage: Headway — Staff Product Designer, Provider Practice Growth
The posting describes an "AI front desk and scheduling assistant" that acts on behalf of providers with prospective patients. The failure is a system that handles consequential interactions — someone seeking a therapist — without earning trust from both sides of the workflow.
Line 1. Prospective patients fall out of providers' fragmented intake and scheduling workflows. Providers are operating a digital front desk instead of delivering care. The posting calls this one of Headway's largest growth levers.
Line 2. Alle (BCG Digital Ventures) — designed a dual-surface system where providers and consumers needed different things from the same workflow, which required defining automation boundaries and trust signals for both sides. That maps because the AI front desk is a two-sided problem: the provider needs to trust what the system does on their behalf, and the prospective patient needs to trust the interaction they're having.
Line 3. When the AI front desk could improve patient conversion but creates a higher risk of an inappropriate or incorrect interaction, who chooses the automation boundary and who can prevent the change from entering production?
Line 4. Downgrade if Design owns the funnel outcome but cannot alter the automation rule, the success metric, or the production decision. Owning the experience without owning the boundary is accountability without authority.
Function-building proof leads here. The posting says you'll anchor a pod and define interaction patterns for a product that doesn't fully exist yet. Show that you've created organizational and product coherence while remaining close to direct work.
Enterprise platform: Atlassian — Principal Product Designer, Rovo & AI
The posting names Rovo, Jira, Confluence, Bitbucket, and "other applications" and says the role creates "a unified cross-portfolio view." When a posting names that many products, the diagnosis writes itself: the fear is that AI capabilities ship incoherently across the suite, and each product team's local decisions compound into a fragmented user experience.
Line 1. AI capabilities and interaction patterns across Rovo, Jira, Confluence, Bitbucket, and other Atlassian products are incoherent. Without a cross-portfolio design authority, each product team builds its own AI interactions and the suite fragments.
Line 2. Alibaba — redesigned an outdated, non-localized B2B platform at enterprise scale, requiring mandate creation and cross-functional leadership to achieve product coherence across multiple surfaces and stakeholder groups. That maps because the core challenge is creating a unified direction across product teams that have their own roadmaps, priorities, and reasons to resist standardization.
Line 3. What is a recent case where this role or its closest peer changed a cross-product roadmap or required a team to adopt an AI interaction standard it had not initially chosen?
Line 4. Downgrade if "unified view" means publishing guidance and coordinating design reviews while individual product teams retain unrestricted authority to ignore the standard. That's a consulting engagement with a principal title.
Published case studies with business outcomes lead here. Enterprise hiring committees evaluate at scale — they want to see that you've operated across organizational complexity, not just product complexity.
Healthcare/regulated: Maven Clinic — Senior Staff Product Designer, Care Delivery
The posting says the designer will "define how AI communicates uncertainty and defers to human care." That language is Line 1 in plain text. The company is afraid the AI won't know when to stop, and the wrong threshold produces clinical harm.
Line 1. AI behavior in virtual-care and provider experiences mishandles uncertainty, human deferral, or emotionally charged healthcare decisions. Without this role, AI interactions in care delivery lack consistent confidence indicators, handoff patterns, and deferral rules, and the consequences are clinical.
Line 2. Thermo Fisher mySupply (BCG Digital Ventures) — exception-first triage in a regulated enterprise environment where the system had to surface the right information at the right moment for a consequential decision. That maps because Maven's care-delivery AI faces the same structural problem: the system must communicate what it doesn't know and defer to human judgment at the right threshold, in an environment where getting the threshold wrong has real consequences.
Line 3. When Design, Clinical, Product, and Engineering disagree about whether an AI interaction communicates enough uncertainty or should defer to human care, who has final authority and what can this role change before launch?
Line 4. Downgrade if the designer is accountable for trust and the member/provider experience but cannot alter the model output, the deferral rule, the evaluation criteria, or the release condition. The posting says you'll "create momentum without direct authority." Line 3 determines whether that's collaborative influence or structural powerlessness.
Published case studies in regulated or high-stakes environments lead here. The evaluator needs to see that you've operated where errors have consequences beyond user frustration. Thermo Fisher carries that. American Red Cross can support — consequential casework, system consolidation under operational pressure — but don't position it as healthcare experience.
Ready to contact vs. still researching
Ready to contact. All four lines filled with specifics. Line 1 names a failure. Line 2 names an artifact and explains the transfer. Line 3 contains a question about a concrete decision right. Line 4 names a structural condition. No line contains language that could apply to thirty companies unchanged.
Still researching. Any line that contains a generic statement or can't be completed. When that happens, name the gap:
- Can't complete Line 1? You don't understand what breaks if this role stays empty. Go back to the posting and its adjacent postings.
- Can't complete Line 2? You can't articulate why your evidence transfers. The evidence might not transfer — that's useful information. Pick a different lead or reconsider the target.
- Can't complete Line 3? Your question is too generic to produce a diagnostic answer. Rewrite it until it asks about a specific decision involving a named tension at this company — then read the stall-point guidance below.
- Can't complete Line 4? You probably haven't done the authority analysis yet. Do that first.
Line 3 is the line most candidates get stuck on. Accept that it's a question you're bringing to the conversation, not an answer you need before the conversation. If it asks about a recent decision involving a named tension — quality vs. speed, automation vs. safety, standardization vs. team autonomy — Line 3 is complete. If it asks about philosophy or culture, rewrite it until it asks about a decision.
Four lines filled, you contact. Any line incomplete, you know exactly what's missing.
- Player-coach means several jobs: OpenAI's Codex growth lead, Anthropic's Design Systems role, and Gusto's Senior Staff Finance seat all advertise hands-on leadership, but the pulse check's cross-tier findings show the phrase can mean preserving senior judgment, eliminating a design-engineering handoff, or compensating for insufficient team capacity — and your Line 1 should reflect which one.
- Staff IC carrying Director authority: Atlassian's Principal Rovo & AI role and Gusto's Senior Staff Finance seat both carry cross-portfolio scope comparable to many Director mandates, which the Atlassian posting makes visible through its strategic vision, coaching, and cross-product ownership language — worth checking whether your Line 4 accounts for title-scope mismatches in either direction.
- Ambience's governance model as Line 3 material: Ambience's published Chorus product description names clinicians, health systems, and the Clinical AI team as holders of review and enablement authority without disclosing what Design can change or stop, making it a useful reference for how to read a company's product materials for authority gaps even when the role itself is no longer posted.
- Shared-information bias in panels: Research on hiring-committee deliberation found that groups disproportionately discuss information shared by all members rather than distinctive evidence held by one interviewer, which means your Line 2 evidence choice needs to survive retelling by someone who saw it for twenty minutes, not just impress the person you showed it to.
