R/reports/design-product-ux

Must have

• Figma, FigJam
• Claude Code, Cursor, Claude, ChatGPT
• v0, Framer, ProtoPie, Principle, Origami Studio
• HTML, CSS, JavaScript, React, TypeScript

Nice to have

• Figma Make, Loveable, Replit, AI Studio, AntiGravity
• Storybook
• Adobe Creative Suite, Photoshop, Illustrator, Sketch, Blender, Maya
• Jira, Looker, Salesforce, Slack, Google Apps, Google Slides, Chrome Console
• Git, Tailwind, Swift, Jetpack Compose

Designing AI into the product

• Human-in-the-loop workflows - deciding when to surface AI vs require human judgment, override/approval checkpoints, escalation routing between automated and human support, preserving user agency when the model is wrong.
• Probabilistic-system UX - confidence levels, uncertainty representation, fallback states, graceful degradation, failure-mode and error patterns for high-stakes decisions.
• Agent and conversational interfaces - chat/voice surfaces, agent personas and interaction patterns, tool-calling and intent-based navigation, AI observability, multi-step agentic flows.
• Trust and grounding - citations/grounding outputs in real data, transparency, fairness/privacy principles, AI quality rubrics, eval signals, human-in-the-loop review loops.
• Mental models for AI - helping users understand, trust, and steer model output; defining "great" for automated tooling in a category with no established conventions.

AI-assisted design practice

• Prototyping in code - Cursor / Claude Code / v0 to produce production-quality UI, design-to-code handoff, "vibe coding" to validate with real interactions not static mocks.
• LLM-driven discovery - using LLMs to clarify problem spaces, draft PRDs/briefs, surface edge cases, pressure-test solutions, synthesize cross-functional inputs.
• Reusable AI workflows - prompt libraries, component generators, eval pipelines, workflow automations that raise team velocity (senior/staff bar).

Design systems

• Scaling and governance - components, tokens, variants, documentation, migrations, multi-surface adoption, enabling non-designers to self-serve.
• AI-native systems - machine-consumable schemas/metadata so AI can ingest the system, unified AI pattern libraries, refactoring legacy UI into a structured source of truth.

Systems and platform thinking

• Primitives and abstractions - data connectivity, permissions/RBAC, authentication, configuration-vs-convention tradeoffs, composable patterns reused across product surfaces.
• Information architecture - navigation, search, content modeling, schema explorers, drill-downs, making dense/data-heavy interfaces legible.
• Cross-product coherence - keeping admin/settings/billing/identity consistent as the product footprint grows; service blueprints across handoffs between product, ops, and AI.

Product strategy and 0-to-1

• 0-to-1 creation - problem framing, MVP definition, defining objects/rules/states/exception paths, roadmap influence.
• Storytelling - narratives that align cross-functional partners and leadership around a vision and tradeoffs.
• Outcome ownership - tying design to measurable business metrics (activation, adoption, retention, conversion, contact-rate reduction), defining success metrics with PM/Eng.

Research and data

• Mixed-methods research - qualitative + quantitative, usability testing, task success and time-on-task, synthesizing telemetry/support/sales signals.
• Data-informed iteration - funnel drop-off and retention analysis, A/B experimentation, post-launch iteration.

High-trust and technical domains

• Regulated/high-stakes UX - payments, risk, fraud, trust and safety, compliance, identity verification, onboarding; translating tax/regulatory rules into clear flows.
• Developer and technical product UX - APIs, SDKs, documentation, console/CLI/IDE workflows, infrastructure and container platforms.
• Multi-surface and real-time - native iOS/Android, web, kiosk, voice/streaming/latency-sensitive, multimodal, console/VR.

The 20% that covers 80%
Across nearly every posting, two clusters gate the role: hands-on AI-in-the-product design and AI-accelerated workflow. Master these first and you are credible for the majority of listings.

• Claude Code / Cursor for prototyping in code
• Human-in-the-loop, uncertainty, and failure-state patterns
• Figma + a mature design-system practice

Table-stakes vs differentiators
Figma fluency, design-systems contribution, end-to-end ownership, and user research are assumed everywhere - they no longer set you apart. The differentiators are shipping LLM-driven product experiences with measurable impact and prototyping in production code.

• Differentiator: designed probabilistic-system UX (confidence, fallback, escalation) in a high-stakes domain
• Differentiator: built reusable AI workflows/prompt libraries others adopted

Fastest path to "qualified"
The smallest high-leverage cluster: systems thinking + AI-in-product + code-based prototyping. One portfolio piece showing a complex workflow made simple, with AI surfaced responsibly and a prototype you built in Cursor/Claude Code, satisfies most postings at once.

Rising vs fading
Rising: agentic/conversational UX, designing with AI as a "core material," design-to-code, AI-native design systems. Fading as standalone signals: pure visual polish, static-mock-only portfolios, "responsive web" without true native or multimodal depth.

What transfers across domains
The high-trust/regulated skill set (payments, risk, compliance, identity) recurs at Stripe, Gusto, Rippling, Brex, Mercury, Plaid, DoorDash Integrity - one investment unlocks the entire fintech/B2B-platform cluster. Developer-tool fluency unlocks Docker, Vercel, Together, Fireworks, Retool, Anthropic, Stripe internal tools.

Interchangeable vs non-negotiable
Interchangeable (learn one): Framer / Principle / ProtoPie / Origami for motion-prototyping; Claude / ChatGPT for LLM discovery. Non-negotiable: Figma, plus demonstrated Claude Code or Cursor use - several postings name them explicitly and test for them.

Gaps that create leverage
Frequently required, rarely held well: representing AI uncertainty and designing graceful failure; machine-consumable design systems for AI ingestion; genuine code-based prototyping (not just "design-to-dev handoff"). Strength here is rare and immediately visible.

Company-stage signals

• Growth (Semgrep, Deepgram, Docker, fal, Decagon): builder mentality, define "good" from scratch, breadth across surfaces.
• Unicorn (Stripe, Gusto, Rippling, Faire, Brex): systems/platform scope, 0-to-1 inside complexity, cross-team influence.
• FAANG-plus (Google DeepMind, Netflix): deep motion/interaction craft, multimodal, behavioral data, formal portfolio fluency in AI patterns.

Lead with measurable AI-product impact
The single strongest signal is "shipped an LLM-driven experience to production with a number attached." This is what unicorns and AI-first companies screen for first.

• Strong: "Shipped Gus AI assistant flows; cut task completion time 40% and reduced support contacts 25%."
• Weak: "Explored how AI could improve the experience."

Show the prototype was built in code
Postings repeatedly distinguish "designs with AI as a core material" from "uses AI as a talking point." Name the tool and the shipped artifact.

• Standout: "Prototyped and shipped onboarding flow in Cursor/Claude Code, iterating with engineers on production UI."

The immediate-yes combo
Complex/regulated domain + AI surfaced responsibly + systems-level solution. This reads as senior and rare.

• Example project: a fraud/risk or payments flow where you designed confidence indicators, human-override checkpoints, and escalation routing, then measured trust/adoption.

The "wow, that's deep" combo
A design system that machines can consume - tokens/components plus schemas and documentation an AI can ingest and generate from. Very few designers have this.

• Example: "Restructured a fragmented UI library into a single source of truth with machine-readable metadata enabling AI-generated components."

Projects eng managers are hoping to find

• Agentic console UX: build/manage/observe AI agents across chat and voice → "Designed the agent-builder and observability surface adopted by N enterprise customers."
• Plain-language-to-answer data UX: question → validated answer grounded in a data graph → "Designed query/schema-explorer flows that let non-analysts get trustworthy answers."
• High-trust 0-to-1: a new payments/benefits/retirement offering from problem framing to shipped MVP → "Led 0-to-1 design and launch of a new benefits product; defined objects, rules, states, and AI-assisted decision moments."

Reads junior
Screen-by-screen execution, visual polish without systems or outcomes, "responsive web" only, AI mentioned but not in the product or the process.

• Avoid: "Redesigned the dashboard UI in Figma for a cleaner look."

Reads senior - same work, leveled up
Frame ownership of strategy, systems, and cross-team outcomes.

• Senior: "Owned end-to-end dashboard redesign across 4 teams; defined reusable configuration primitive (no-code), aligned PM/Eng on the roadmap, and shipped a 30% activation lift."

Red flags that make recruiters pass

• "Familiar with AI tools" with no shipped AI feature or workflow.
• Portfolio with no quantified impact (several postings will not consider a portfolio without it).
• No design-system contribution shown.
• Avoid: "Passionate about AI and emerging tech" as a standalone line.

Open observation
Many roles (Gusto, Rippling, Ramp, DoorDash) now expect designers to "design in code" and contribute to implementation - front-end literacy (React/TypeScript/CSS) is shifting from bonus toward baseline for senior/staff. A repo or coded prototype alongside the portfolio is increasingly a distinguishing signal rather than an engineer-only credential.