Somebody writes down what you are before they read a single case study.
Everything below rests on that, so start with the evidence. Two companies publish their design hiring process in enough detail to be worth reading. Atlassian runs a thirty-minute recruiter screen on background, skills, and career aspirations, then a hiring-manager call built on structured scenario questions, then a portfolio review as a separate session later. Wealthsimple runs the same shape: recruiter conversation, hiring-manager conversation about past projects, portfolio review with product, design, and engineering in the room.
Count the conversations. You describe yourself twice before you show anything.
And the description travels. Greenhouse scorecards, the structured forms most companies record interviews in, capture a conclusion, pros, cons, and a pass-or-advance recommendation. Earlier feedback can be surfaced to the hiring manager during review. Executive search has the same plumbing: AESC's standard for member firms puts interaction notes in the longlist report and competency-scored interview reports in the shortlist.
High confidence that the early notes are structurally available downstream. Moderate confidence on how often they actively steer the later read, because nobody publishes that number.
Still clear enough to act on. By the time your portfolio opens, it is being read as evidence for a label someone already wrote down. Which sharpens something I got slightly wrong in Where You Lose the Room, Gate by Gate. I treated the recruiter screen as a gate you survive. Wrong altitude. It is the gate where your species gets recorded and forwarded.
So leading with a published design decision belongs in your answer to "tell me about your background," spoken long before any deck opens.
Four rooms, four withheld identities
Head of Product is not one obstacle. It reads differently depending on who is holding it, and the four archetypes withhold the design-leader identity for four unrelated reasons:
- AI-native suspects you no longer make.
- Growth-stage platform suspects you will delegate instead of ship.
- Enterprise platform suspects you have never run a function through a performance cycle.
- Healthcare and regulated suspects you have never carried the cost of a wrong answer.
The logo will not tell you which doubt is live. An AI-native company hires an enterprise-flavored design VP and inherits her tests. A growth-stage company that starts selling into hospitals develops consequence anxiety. Diagnose in the room, off their first real question:
| Their first real question | What it tests | Whose doubt is live |
|---|---|---|
| "Who did design report to, and how big was the team?" | Function ownership | Enterprise |
| "How involved were you in the actual design work?" | Making | AI-native or growth-stage; the follow-up about timelines tells you which |
| "What happened when the system got it wrong?" | Consequence | Healthcare grammar, whoever is asking |
Then run the same three beats in that room's dialect.
- Beat one: a specific design decision from your published work at junochen.com, described at the level of mechanism.
- Beat two: TinyFish, entering only as the reason that particular judgment got more precise.
- Beat three: the career narrative, once the first two have landed.
All of this is spoken. It is what you say when asked, not what you put on a slide. The deck is a separate problem with separate rules.
Order carries the weight here, not content. Evidence has to land before the label, because after the label lands the evidence gets re-read to confirm it.
Why you cannot just say it
You cannot assert the identity into existence, and the rooms hardest to convince are exactly the ones already watching for drift.
The cleanest evidence comes from medicine. A study of NHS Scotland consultants found claimed leadership identities routinely refused by peers, and it named the refusal mechanism: colleagues judged that the hybrid had moved too far from the profession he now wanted to lead. Seventy-seven percent said they wanted more clinicians in senior leadership. Thirty-two percent thought the clinicians who got there were any good at it. Wanting the category to exist and granting it to a named person are two different acts, performed by the same people, on different criteria.
No designers in that study, no portfolios, no hiring panels. Call it suggestive. What transfers is the mechanism: identity gets granted through observed behavior, never through declaration. So does the specific shape of the resistance you will meet.
A second fact about your position should change your opening move more than it looks like it will. Nobody counts your path. The Nordic State of Design 2025 survey ran over a thousand responses and does not report prior titles for Director-plus hires. A 2026 product management survey found 14% of product respondents came from design, which measures traffic in the other direction. Senior precedent exists and is public. Chris Abad spent roughly a decade in Head of Product roles before returning fully to design and going on to lead design at Dropbox, Square, and Google. But there is no benchmark and no count.
Recognized, unscripted. If your move were common you would be arguing against a stereotype, and rebuttal would be the instrument. It is not common. Most panels hold no stored template for you at all. Nothing is being rejected here; something is being invented, live, out of whatever you hand them first.
Hand them a decision, not a trajectory.
AI-native, where they suspect you no longer make
What they scan for. Hands. Specifically, whether you can still reason about an interaction at the level of the mechanism rather than the outcome.
How the title lands. Better here than in any other room. These companies are writing product scope directly into design roles, and production exposure to agent behavior is thin on the market. Moderate confidence, inferred from how the roles are written rather than measured.
Beat one, out loud. One decision from your Agentic AI work, pitched at this altitude:
"I gated the write action behind a reversible preview instead of a confirmation dialog, because the confidence signal the model exposed was well calibrated on retrieval and badly calibrated on action, and a confirm step just transfers a decision the user has no basis to make."
Mechanism, rejected alternative, failure mode. That is a designer talking.
Beat two. TinyFish in one sentence, as the thing that made that calibration more precise, and name the change specifically. Keep it small and verifiable.
What gets you killed. Outcome language with the mechanism stripped out. "We improved trust in agent output" is a product summary and it ends the conversation. The other killer is the passive team construction. Every "my team explored" is a deposit into the suspicion.
Exception cue. Some AI-native roles run the making test late instead of first. Altana's Head of Product Design posting opens on building an agent-first design organization: structure, metrics, hiring, rituals, business outcomes. Working prototypes and code fluency appear further down. The test moved. It did not disappear. When a posting opens on org architecture, do not disarm.
Growth-stage platform, where they suspect you will delegate
What they scan for. What you did the last time the work was late and the quality was wrong. That answer decides the room, not your opener.
How the title lands. Badly, and quietly. Head of Product is a directing title. Growth-stage roles are player-coach, meaning the leader is expected to keep their own hands in the work, and directing is the precise failure mode being screened for.
Beat one, out loud. A decision you made yourself, with the clock audible in the sentence:
"We were eight days out and the onboarding flow had three states we couldn't get right, so I cut the third one myself over a weekend and shipped the degraded version with an explicit empty state rather than slipping the date. I was wrong about which of the three mattered, and we rebuilt it in the next cycle."
Deadline, your own hand on the work, a tradeoff you absorbed instead of escalated, and a cost you paid. Your 0→1 pillar is where this lives.
Beat two. TinyFish enters as the reason your judgment about what to cut got faster. Stop there.
What gets you killed. Roadmap grammar. Quarters, bets, prioritization frameworks, stakeholder alignment. The dead phrase is "we aligned the team on the tradeoff": true, correct, and a description of a meeting rather than a decision. Answering a craft question with a process answer is the same mistake in a different suit.
Exception cue. Front's Head of Product Design role is unusually altitude-heavy: managing managers, operating cadence, governance, executive communication, design-system leverage. It is the closest thing to a pure strategic growth-stage role in the set I reviewed, and it still says hands-on. At growth stage, higher-altitude language is additive. If you think you have found a posting that traded player-coach for management, read it again.
Enterprise platform, where they suspect you have never run a function
What they scan for. Not making. Whether you have held a quality bar across a team through a performance cycle, leveled people against it, defended headcount for it, and kept a system coherent while the org reorganized underneath you.
How the title lands. It supplies scope and withholds function ownership. Head of Product tells them you had influence. It does not tell them a design organization ran on your calibration.
Beat one, out loud. From your Alibaba enterprise pillar, make the object of the decision a rule that other people applied. The shape you want, in your own facts:
"We were getting inconsistent provenance treatment across surfaces, so I wrote the rule that any result whose ranking was model-influenced had to carry its source inline, not on hover, and then I sat with three teams to see where the rule broke. It broke on dense comparison views, so we changed the rule rather than granting exceptions."
You decided, other designers executed, and you knew when their work cleared the bar. That last part is the enterprise proof. Deciding well is table stakes at this level; making your judgment reproducible by people who are not you is what is being tested.
Beat two. Hold TinyFish. This is the one room where volunteering it costs you, because it invites the panel to picture a product org instead of a design function. Answer it when asked, frame it as current technical currency, return immediately to what you ran.
What gets you killed. "I didn't have formal authority, but I had a lot of influence." Influence is the enterprise operating condition, not an achievement, and describing yourself as effective without a team confirms the exact suspicion in the room. Say what you ran and who reported to you.
Exception cue. Salesforce's Director, User Experience posting asks for five-plus years managing design teams and the ability to write code for high-fidelity prototypes, with coded artifacts in the portfolio and recent AI building. When an enterprise posting names design engineering or agentic tooling, the test became builder-plus-manager. The requirement was added on top; nothing came off.
Healthcare and regulated, where they suspect you have never paid for a wrong answer
What they scan for. Consequence literacy. They listen hardest at the moment you describe a tradeoff, for whether you know what happens downstream when a system is confidently wrong and a licensed professional acts on it.
How the title lands. Partly in your favor. Production AI exposure is scarce, and regulated buyers are under real pressure to ship AI features they do not yet trust. That is genuine currency in the wrong denomination for the test being run. No title confers consequence literacy.
Beat one, out loud. From your Agentic AI trust essay, the override, the escalation path, the audit trail. Say the human's name out loud in the sentence:
"The person signing that record is personally accountable for it, so the design question was not whether to show confidence but what she needs in front of her at 2 a.m. to justify a reversal to someone reviewing it six months later. We surfaced the disagreeing evidence, not the score."
Name who bears the consequence. Name what you handed them to bear it with. That is the Crumple Zone Test from Battlecards, the question of who absorbs the impact when automation fails, turned on your own credibility.
Beat two. One sentence: TinyFish is where you watched production agents fail in ways demos do not, which is why you now design for the wrong answer first. Then hand the conversation back to consequence.
What gets you killed. Speed as the headline. "We cut the workflow from six steps to two." In this room that sentence identifies someone who has never watched a fast wrong answer reach a patient.
Exception cue. Jukebox Health's senior design posting opens on taste and point of view and asks for polished interfaces in clinician and health-plan workflows. Craft leads; domain consequence is context. A regulated posting that opens with taste rather than compliance is a company buying design maturity it does not have in-house. Lead with craft. Hold consequence in reserve and produce it when they test for it.
Beat three, and how to say the return
Beat three is not optional and it is not filler. It is the language you use after the decision has landed. The tell that they are ready for it is that they ask directly. "So how did you end up in product?" That question means the design read is provisionally granted and they are reconciling it, which is a good problem to have.
Public accounts of designers who went into product and came back sort into two language families. The phrasings are attributable, the classification is mine, and I have no evidence on how panels respond to either. High confidence on the words, low on the landing.
Accumulated capability. Julie Froelich has written that product management made her a materially better designer and leader. Yuhki Yamashita, who now runs both functions at Figma, describes oscillating between the two and argues the boundary should be blurrier than it is.
Affinity and distance. Abad describes product pulling him away from what he enjoyed. Prarthana Johnson credits her return to love of the craft. Aly Blenkin, moving back to an IC role, describes missing the work after management became slide decks.
The second family is honest and human. In a Director-plus conversation it also describes someone whose reason for being in the room is preference, and preference does not argue for scope. Use the first family.
The first has its own trap. Abad's framing that a step backward can be a step forward concedes a hierarchy nobody in the room had applied yet. Never introduce backward. If they have not framed your move as a demotion, do not do it on their behalf.
The worse failure is oscillation. A 2026 coaching account drawn from consultations with displaced senior leaders describes candidates re-cutting their positioning between senior IC and strategic director inside a single process, leaving hiring teams unable to say what the person is. No sample size, so: practitioner observation. It matches what I watched in search after search. Two interviewers write down two different candidates. The debrief then burns its forty minutes reconciling the discrepancy instead of assessing the person, and the room resolves the discomfort the way rooms always resolve it, by advancing the weaker candidate everyone could describe in one sentence.
Pick the version. Say it the same way to the recruiter, the hiring manager, and the panel.
The ledger
Two rooms where the title is currency. Two where it is a tax you pay down in the first ten minutes.
Currency, ranked. AI-native first, decisively. The product-leadership scope is precisely what these companies are writing into design roles, and production agent exposure is scarce. Healthcare and regulated second, narrowly. Your AI fluency buys attention, not the identity. The identity still comes from consequence, and only your published work proves that.
Tax, ranked worst first. Enterprise platform, because the title supplies business fluency against a test that is explicitly about running a design function and holding a bar. Then growth-stage platform, because a directing title signals delegation into a role built around shipping.
The rule from the Buyer-Angle Map in Portfolio Playbook survives all four rooms unamended. TinyFish is currency, never collateral. It explains your title and sharpens your judgment. Your published work is the only thing that proves you are a designer.
Lead with the decision. The career story is beat three for a reason. Told after the evidence, it explains a designer who went and got something. Told first, it is the only thing anyone remembers.
- Read the org-building exception closely: Altana's Head of Product Design posting is the cleanest current example of an AI-native company asking for structure, hiring, rituals, and metrics before it asks for prototypes, and it is worth reading in full to see where the making test reappears.
- Watch the enterprise builder shift: Salesforce's design-system transformation is now partly owned through Product Management roles rather than a design reporting line, a signal worth tracking against its Director, User Experience posting when you assess where authority actually sits.
- Check the explanation research before you use the word trust: A 2021 CHI study of cognitive forcing functions found that the interventions that most reduced overreliance on AI were the ones participants liked least, which is a sharper opening argument in an AI-native room than any explainability claim.
- Test the identity-granting mechanism yourself: The 2026 Annual Review on career transition and professional identity synthesizes 133 studies on nonlinear paths, provisional selves, and liminality, and it is the most useful single read for understanding why the panel invents your category rather than retrieves it.