Juno, your "Trust Is the New Interface" essay is a positioning instrument. The five-handoff framework maps to the dominant anxiety in design leadership postings across tiers right now. Every company writing a Director+ role description in July 2026 is circling the same problem with different language. Your essay gives them the language. Your portfolio proves you've shipped at every seam they're worried about.
The framework gives you shared vocabulary with nearly every buyer in the current market. Your published cases prove you've shipped these handoffs in production, not theorized about them.
Organized by handoff, not by company. All portfolio cases live at junochen.com.
Intent-Setting
Confidence: high. Cleanest match between essay framework and live buyer language.
The essay puts the first design problem before the agent acts. Human states intent. Agent interprets. The gap between what was meant and what was understood is where trust breaks first.
Amplitude's Head of Product Design posting describes this gap without naming it. They're building an AI-native platform where "agents and humans work together" in shared workflows. The stated design leadership problem: maintaining coherence and quality when AI makes UI changes cheap to ship. Cheap to ship means cheap to misinterpret. That velocity is exactly where intent-setting failures propagate.
Two cases prove this. Alibaba's B2B redesign started with the structural finding that 25% of traffic generated 80% of transaction value, but high-value buyers lacked procurement-grade trust signals. The agentic vision you published closes with a natural-language procurement brief where the buyer states intent and the agent traverses search autonomously. Intent-setting design at transaction scale. Thermo Fisher adds the upstream version: 32 warehouse-floor interviews became five failure modes became five modules. You designed the translation layer between what operators needed and what the system could interpret. Intent-setting before anyone used the word "agent."
In-Progress
The essay's second handoff is about visibility while the agent works. An active agent that goes silent becomes unreadable. Silence is the fastest way to lose a user's trust.
Salesforce's Lead Product Designer, Agentforce states the in-progress anxiety more explicitly than any other current posting. The role defines patterns for how agents "communicate, hand off work, use memory, and coordinate in real time." That phrase, "coordinate in real time," is the in-progress handoff stated as a design requirement. Salesforce wants interaction models scaling across Salesforce, Slack, and the broader platform with "reliability, intentionality, and value across workflows." Reliability and intentionality are trust words. They are asking for what your essay argues for.
Your proof: Brand Pulse monitors Reddit and X in real time, scores mentions for sentiment, urgency, and source, and turns live signals into a decision surface. The agent is always working. The design problem was making that work legible, inspectable, interruptible. Thermo Fisher's capacity heatmap does the same in a physical-logistics context: constraints and performance gaps become visible before commitments or delivery failures, not after. In-progress visibility at its core is a trust surface. You earn the right to keep the agent running by showing the human what it's doing.
Output Review
Third handoff: the agent presents its work and the human evaluates it.
"Accuracy is what the model achieves; consistency is what the design delivers."
The model may be right. The design must make rightness verifiable and wrongness detectable.
Gusto's Senior Staff Product Designer, Finance states this with unusual precision. The role requires designing AI-first financial UX that "represents uncertainty clearly, surfaces errors gracefully, preserves agency when AI recommendations are wrong." Designers must "critically evaluate AI-generated outputs and know when to override or course-correct." Gusto's broader design philosophy frames this as the question of who owns AI wrongness and exception movement across systems and teams. Language varies by posting. The anxiety is identical.
Strongest output-review proof: Carrier IQ. Insurance quotes with confidence scores. A side-by-side broker decision surface. The broker trusts the quote because the design made the basis for the quote inspectable, not because the system said to. Allē adds a second proof point: the consumer surface moved from static balance display to progress milestone, the provider surface added patient context at checkout. Both are output-review designs that give the human enough information to evaluate what the system produced.
Decision Gate
Confidence: high. Deepest portfolio match. The handoff where your essay draws its hardest line.
"Automate what you can verify. Keep a human where the outcome is irreversible."
Headway's Design Director, Provider Experience describes a product surface where decision gates are clinical and financial simultaneously. The provider EHR touches scheduling, telehealth, clinical documentation, outcome measurement, insurance verification, and billing within a single session scope. Those workflows are being "reimagined with AI at the center," including transcript-powered session notes and AI-enabled scheduling. Every one contains at least one irreversible moment: a clinical note submitted to an insurer, a billing code attached to a patient record, a treatment outcome logged. The design problem is where the human must verify before the record becomes permanent.
Your decision-gate evidence is the deepest in your portfolio. Thermo Fisher shows that exceptions surfaced at the delivery gate were 3–5x costlier to fix than earlier detection. The agentic vision preserves one irreplaceable human gate: batch QA release requiring regulatory signature. Red Cross reinforces through a different domain: one shared record across role-specific views, eligibility checks, identity verification, FEMA audit-trail generation, and money movement under federal oversight. The decision gate is trust architecture. You've built it twice, in two regulated contexts.
Loop Feedback
Confidence: moderate. The anxiety is real but distributed across postings rather than concentrated in one.
The essay's fifth handoff addresses what happens when the agent runs again. Output from one loop becomes input for the next. Small errors compound across repeated runs.
Amplitude's concern about coherence when changes are cheap to ship is a loop problem: each shipped change becomes context for the next agent action. Salesforce's "agent memory" language is a loop problem: what the agent remembers shapes what it does next. Neither posting isolates this as a standalone requirement. Both embed it in their broader coherence anxiety.
Loop-feedback proof comes from Brand Pulse and Retail Velocity, and the specific design decisions in both are anti-compounding measures. Brand Pulse publishes hourly brand-health scoring and a weekly agent-written narrative. The hourly cadence is the design choice: it makes drift visible before it compounds into a weekly narrative that misleads a CMO. Retail Velocity runs daily account audits with "what changed" prompts that force the field rep to confront yesterday's output before acting on today's. Both are designs where the system's memory is visible, its drift is detectable, and its compounding errors are catchable because the design created inspection points at the frequency where errors are still small.
The Anchor Effect
Each handoff appears, by name or by anxiety, in current live role descriptions. Each one connects to published production work at junochen.com.
Most candidates at this level arrive with strong portfolios and no framework. You arrive with a framework that makes the portfolio legible. More than that: the essay pre-frames how the buyer processes every case you show afterward. Anchoring effects in evaluation panels are well-documented. The first interpretive frame a hiring committee receives shapes how they weigh everything that follows. Your essay sets that anchor. By the time they open a case study, they are already looking for the handoff. They find it.
Lead with the essay. Let the handoffs do the routing.
Three roles in the current market don't fit this framework. Next piece identifies them and routes the proof differently.

