From the Tools section of Will (Xiaohongshu UX/AI job search blogger)'s course. Only the transferable parts are distilled here: how to structure prompts when using a custom GPT (his is called PitchPro, essentially a writing bot that "takes your resume + JD + portfolio screenshots → outputs customized portfolio copy and word-for-word interview scripts"), plus a set of common interview question bank and prompts for telling projects as stories. Product-specific operations (how to top up, how to subscribe to his bot, which VPN to install) are all skipped — that part is onboarding, which becomes invalid when switching tools.
Significance for PMs: ① The bot usage principles are actually the universal skill of "how to collaborate with any LLM to produce high-quality long documents" (drilling down details layer by layer / saving content in a notepad / editing instead of continuing the chat), which applies the same to writing PRDs, retrospectives, and competitive analyses. ② The interview question bank is a ready-made self-check list — use it to ask yourself whether you can answer for each project, and you will know where your explanation is unclear. ③ The creative narrative prompts solve the common problem of "flat project storytelling", and can also be used for performance reviews/pitch presentations. ④ The structural template for interview presentation (Problem→Process→Impact) and STAR, as well as focusing on culture rather than company names are covered in Google UX Certificate / Portfolio · Resume · Interview · Job Search; that piece is the skeleton, this one is the question bank + scripts, use them together.
1. Bot Usage Principles (Universal for Writing Long Documents with Any LLM)
Why this section is listed separately: Most people fail at using AI to write portfolios/interview scripts not because the tool is bad, but because their way of feeding context is bad — a single line "help me sort out the process" won't get you usable output. The 7 points below are the collaboration habits that can truly be transferred to any writing scenario, after stripping away "which button to click" from his product operations.
1. The minimum three-piece set for feeding context: materials + target position + clear instructions
The most effortless way to feed context when writing customized self-introductions/interview scripts:screenshots of your own resume + JD of the target position + a prompt that clearly states "what you want", throw them all in together. The bot will cross-reference to produce a customized draft of "your experience × the abilities required for this position", instead of a generic template.
Reusable template:这是我的简历[截图] + 目标岗位 JD[截图]。请按[指定结构]生成一份[场景:可口述的自我介绍 / 2 分钟项目口述 / ...],只输出正文,不解释。
Why: The LLM doesn't know where you are applying; give it a position anchor so it can do "matching" instead of "listing". This is the same logic as giving business background + target users first when writing a PRD — context determines the relevance of output.
2. Build the big framework first, then drill down into details layer by layer (don't expect to get it right in one go)
The first output from the bot will only be a rough framework like "background → problem → challenge → solution → data → retrospective"; a correct framework doesn't mean it's usable. The correct rhythm:confirm the framework is correct → put forward specific expansion requirements for a certain section → drill down further. For example, if the solution is not specific enough, follow up with "under these three strategies, add an objective verification process and method for each"; if the challenge is not credible enough, follow up with "add a tortuous failure experience to the first challenge to increase difficulty and authenticity". Why: It is impossible to get a fully satisfactory result from AI in one go. Drilling down details in layers = you always control the big structure and only let it fill in the blanks, leading to higher quality and controllability. This is equivalent to "defining the information architecture first before filling in content" in product work.
3. The more specific the instruction, the better — give the subject, purpose, and expected output format
This is the easiest pitfall to step on, so it is emphasized separately. The most feared instruction looks like this: "Write a deduction and analysis framework based on the project" — missing subject (what to deduce), unclear output format (is the framework a chart? A list? A flowchart? A report outline?).
| ✗ Vague and empty | ✓ Specific and executable |
|---|---|
| Help me sort out the process | Further refine the methods and reasoning process of each link before "identifying key problems", to make it more objective, complete and persuasive |
| Write a deduction framework based on the project | Around this design project, sort out a deduction structure with five steps [problem definition → user research → data analysis → insight extraction → solution output] |
Formula for good instructions:Specify exactly what it targets + what format to output + the reason I am asking you to do this. Why: AI won't guess the implicit premises in your head. State clearly "to whom, what you want, why", and it won't go off track with free rein. This is universal for all AI collaboration — the same applies when you write prompts for Claude.
4. Open a notepad at the same time, save satisfactory fragments at any time (to prevent full rewrites)
When using the bot,open a Notion / text notepad simultaneously. After the big framework is set,immediately copy and save every satisfactory part into it. Why: In LLM conversations, modifying a small detail often triggers it to rewrite the entire paragraph, wiping out the parts you were satisfied with just now. The external notepad is your "version staging area", allowing you to only merge the essence and not be held hostage by its randomness. After multiple pieces of content are collected, throw them back and ask it to "synthesize an N-minute word-for-word script while following the original content and logic".
5. If you are not satisfied, edit the original instruction instead of continuing the chat
When the output is wrong,go back to modify that original prompt and resend it, instead of continuously appending new conversations afterwards. Why: Continuing the chat will pile up context more and more messily, and errors will be carried into subsequent rounds; modifying the original instruction = correcting from the source, leading to more accurate results. (Analogy: fixing a code bug requires fixing the root cause, not patching over it later.)
6. Reverse-engineer "data—insight—solution" from UI screenshots (complete the evidence chain of the portfolio)
Portfolios often lack evidence for "why this change was made". Throw a UI screenshot of the solution to the bot,specifically ask it to reverse-engineer: what problems exist in this interface, what data corresponds to them, and exactly where the revision should be made. It can give a complete chain of "problems → data → insights → solutions"Data（数据）→ Insight（洞察）→ Recommendation（优化建议）→ 验证方式（A/B 测试）→ 结果. If you feel the solution leans towards "adding features" rather than "optimizing experience", prompt it "the solution approach should focus on the scope of work of an experience designer, respect the original professional content" or "give me several UX directions".
Why: Interviewers value the explicit connection between "insight ↔ decision" the most (see Google UX Certificate / Portfolio · Resume · Interview · Job Search Appendix Item 5). The bot helps you supplement the logic you only realized after the fact into a structured evidence chain, but you must verify the facts yourself — the data is reasonable values it made up; for projects you actually worked on, replace them with real numbers.
7. Practice answering questions with voice, not typing
When letting the bot do mock interviews,you must answer with voice input, not by typing. Why: Typing gives you plenty of time to think and edit, while real interviews require impromptu, fast, clear verbal expression. Voice input can practice instant reaction, language organization, intonation, speech rate and confidence at the same time — these are all dimensions interviewers assess. This transfers to preparation for any oral report/pitch.
Product-specific operations in one sentence: PitchPro is a custom GPT made by Will, which requires GPT Plus + a VPN to use; finished portfolio inner pages can be used with NotebookLM (feed interview scripts saved as txt/pdf into it to generate PPT structure and image/text references for each page) + Figma for rough layout. These are the specific implementations of his toolchain; the principles (1–7 above) hold true for any LLM.
2. Common UX/Design Interview Question Bank (Grouped into 8 Categories + How to Answer)
Why: This is a question bank that can be directly used for self-check.The usage is not to memorize answers, but to ask yourself question by question "which of my projects can answer this, can I answer it with data" — places you can't answer are where your portfolio is not explained thoroughly enough. Use it together with the STAR method in Google UX Certificate / Portfolio · Resume · Interview · Job Search: use STAR (Situation → Task → Action → Result) for behavioral/experience questions; for cognitive/trade-off questions that test your judgment framework, state your opinion + give reasons + give examples.
0｜Common Opening Questions (High-frequency for HR, prepare standard answers first)
Greatest strength / 5-year career goal / what team atmosphere you want to grow in / your understanding of our company or product / why you are interested in the experience designer position / what unique value you can bring / earliest onboarding date / expected salary. How to answer: These are "match + stability" probes. Your strengths should align with the position JD; don't state career goals higher than this position (it makes you look like you're using it as a stepping stone); for "understanding of the company", be sure to research the product in advance. For salary questions, give a range + leave flexibility.
1｜Context: Project Background (Explain clearly what you made and for whom)
Most satisfying project / your scope of responsibility and goals, what you did around the goals / user personas and usage scenarios / which pain points the overall goal solved / unique advantages compared to competitors / user composition and content sources of portfolio projects / project positioning and main selling points / why it is the most valuable project in the portfolio / how design requirements were determined, how priorities were judged / how target users were segmented, what core needs were. How to answer: First use one sentence to clearly state "solved what problem for whom" (equivalent to role + goal + user in case studies, see Google UX Certificate / Portfolio · Resume · Interview · Job Search 7 Elements of a Case Study). For "why this project is the most valuable", tie it to position match, don't just say you like it.
2｜Purpose: Value and Metrics (Tests PM thinking the most, must have numbers)
Core measurement metrics / how design experience is evaluated, standards and frameworks / where requirements come from, how to calculate ROI / which metrics you care about most, how to calculate them / the process of setting goals and metrics / how the implementation logic of each module is reflected in data / how to balance cost and experience / how benefits are reflected, specific data / core metrics of the most satisfying project. How to answer: This is PM's home turf and also designers' weak point.Prepare at least 1–2 quantitative results for each project("task completion rate +26%" "transfer steps −33%"). Being unable to explain metrics = not having thought through value. For ROI/cost trade-off questions, show that you understand "design must be accountable to business".
3｜Way: Design Approach (Tests process, not results)
What different strategies you would use if you did it again / biggest difficulty and solution / which parts you are responsible for when abstracted into a modular structure / a complete process example of improving experience once / the most valuable part of the project / your specific contributions / end-to-end process and your main role / communication methods with the team/clients / how to identify and solve user pain points. How to answer: Interviewers want the process of "how you went from concept to delivery" (corresponding to Google UX Certificate / Portfolio · Resume · Interview · Job Search "adaptability" among the three key interview points). Explain the chain of "identify problem → hypothesis → verify → iterate" clearly, don't just report results. "How would you change it if you did it again" is a signal of reflection ability, you must be able to answer it.
4｜Result: Outcome Evaluation (Attribute metrics to your design)
How metric improvements are attributed to your design, with examples / how you completed the entire project, project management process / actual effects of your optimization strategies / data support after the solution launched, what areas need improvement. How to answer: The difficulty lies in attribution — prove that "my design brought this number", use A/B control or before-and-after comparison. Being candid about "what needs improvement" actually gives you extra points (see Google UX Certificate / Portfolio · Resume · Interview · Job Search Appendix "Treat failures as selling points").
5｜Replay: Takeaways and Retrospective (Reflection ability and honesty)
Looking back, what could have been done better / biggest takeaway / have you encountered difficulties that you couldn't solve on your own / what would you do if you did it all over again.How to answer: Corresponds to the third key interview focus: "learning mindset." For "difficulties you couldn’t solve on your own," don’t force the image of being all-capable; talking about how you sought help or collaborated to solve the problem instead shows maturity.
6｜Cognition & Methodology (high-frequency trade-off questions, testing your decision-making framework)
Core competitiveness of an experience designer / user thinking vs business thinking / views on the prospects of AIGC in experience design / an AI product with great recent experience / how you usually stay aware of and learn about trends / your favorite app, its strengths, weaknesses and improvements / what you would do if asked to design a similar app / how to drive cross-departmental collaboration / how to judge if a company is a good fit for you / how to process feedback from users/clients / How to balance: innovation vs actual needs, aesthetics vs data-driven decisions, long-term value vs short-term demands, brand consistency vs product features, measurability vs user emotion, innovativeness vs user familiarity, research vs fast launch, aesthetics vs functionality / how to make decisions when needs of different user groups conflict / how to choose when business needs conflict with user needs / how to prioritize when resources/time are insufficient / what to do when PMs and developers propose unreasonable modifications / what to do when you have doubts about your own solution.
How to answer: This entire category consists of "trade-off questions with no standard answer". What the interviewer is looking for is your decision-making framework, not taking sides. Template:先承认两边都有理 → 给出你的判定准则(通常落到"看哪个更服务核心目标/核心用户") → 举一个你真实做过的取舍例子. Be sure to prepare "an AI product with great recent experience" and "improvement points for your favorite app" — these two questions are almost guaranteed to be asked, and thinking them up on the spot will come off as empty.
7｜Management (asked only for team lead/senior roles, covering conflict handling and managing up)
What to do when team members have uneven capabilities / how to lead the team to resolve schedule delays / how to handle negative emotions and unhealthy team culture / how to lead the team with limited resources / how to encourage proactive innovation / how to treat star employees / when team goals are misaligned with individual goals / how to tap into members’ potential and foster a sense of ownership / A full set of managing-up scenarios: how to persuade when your superior has a different opinion but your solution better serves users, when your superior wants a fast rollout but you believe time is insufficient, when your superior’s goals are unclear and frequently change, when you disagree with your superior’s management style, when your superior’s evaluation is not objective, when your superior asks you to produce inflated performance reports, when your superior gives unreasonable negative feedback, when your superior does not support your talent development suggestions.
How to answer: Management questions test "focusing on the issue rather than the person + communicating with data/facts instead of emotion." A safe framework for managing-up questions:理解上级的诉求/约束 → 用客观证据(数据/用户反馈/风险)而非对抗去陈述 → 给可执行的折中或分阶段方案. For situations like being asked to produce an inflated report, hold your bottom line but offer an alternative; do not confront head-on, nor compromise by falsifying data.
How to use the question bank to pass in one go: Treat the 8 categories above as a checklist, pick 1 position you most want to apply for,use the bot to predict 20 questions(see "Reusable Prompts" below for instructions) → answer each question out loud by voice → for questions you can’t answer, go back to your portfolio to add supporting evidence. This is far more efficient than rote memorization of generic questions.
III. 10 Creative Prompt Sets for Telling Projects as Stories (for project presentations)
Why: Having a portfolio script that is factually correct but delivered flatly is the most common case of "falling just short of the finish line." These 10 prompt sets rework the same project content into different narrative structures, all centered on "adding hooks to create suspense, so the interviewer keeps listening as if following a TV series."How to use: First have the bot produce a standard 5-minute interview script, copy it into the [Project Brief] position in any of the templates below, and generate a more engaging 10-minute version. This corresponds to Google UX Certificate / Portfolio · Resume · Interview · Job Search Appendix Item 9 "Tell the case as a real experience on a particular day" — turn monologue into dialogue.
General opening:我即将参加专业 UX 面试,想用更有创意的方式讲项目,请基于以下[项目简介],生成时长 10 分钟的面试话术逐字稿:______
| # | Narrative structure | One-line approach | Hook example |
|---|---|---|---|
| 1 | Documentary style | Reconstruct the real process chronologically, with specific time/location/roles/conflicts | "However, the real challenge had only just begun" |
| 2 | Single-take from the user’s perspective | Use the first person of the target user, unfolding emotions and operations along a one-day timeline | "Just when I thought everything was over, something unexpected happened…" |
| 3 | Mystery detective drama | Open with an anomaly/suspicious point, reveal investigation (data/interviews/validation) in stages | Raise a new question at the end of each discovery |
| 4 | Campaign retrospective method | Divide the story into 3–5 "campaigns" covering strategic goals/tactics/execution/results | "The next step will be even harder" "An unexpected obstacle emerged" |
| 5 | Dual parallel narrative | One thread covers execution, the other covers metric changes, echoing each other | "However, the data told us a completely different story" |
| 6 | Failure-reversal | First lay out a failure (background/causes/impact), then tell the reversal | "Just when everyone thought the project would be abandoned…" |
| 7 | Highlight flashback | Open with the most impactful result, then trace back the process | "This seemingly simple result actually went through…" |
| 8 | Bystander notes | Narrate from a third-party perspective (colleague/user/leader) to increase authenticity | "At the time, we had no idea what would happen next" |
| 9 | Future retrospective method | Assume it is one year after project completion, look back at the long-term impact from a future perspective | "And the turning point for all of this actually happened at…" |
| 10 | Micro-detail entry point | Zoom out from a small detail to the big picture, wrap up with data + user feedback | "It was this seemingly trivial detail that triggered…" |
How to choose: If you have impressive data → use #7 flashback or #5 dual narrative (let the numbers become the suspense); if the project hit major pitfalls → use #6 failure-reversal (turn the setback into a selling point); if the process is long and complex → use #1 documentary or #4 campaign (break it into digestible segments); if you want to highlight empathy/experience → use #2 single-take user perspective.Note: Narrative is packaging; the core (problem → process → impact) must not be overshadowed by flashy presentation, and data and insights must remain clear — this is consistent with Google UX Certificate / Portfolio · Resume · Interview · Job Search appendix "complexity gives way to clarity."
IV. Quick Reference Cheatsheet for Directly Reusable Prompts
Why: Gather the effective prompts scattered earlier into one table; replace the gray variables to use.
| Scenario | Prompt template (replace [variables]) |
|---|---|
| Customize self-introduction | 这是我简历[截图]+目标岗位 JD[截图],按可口述的 UX 自我介绍结构生成 45–60 秒话术,只输出正文不解释 |
| Build high-level project framework | 把这份作品集截图当输入,生成"背景→问题→挑战→方案→数据→复盘"的项目框架(don’t make the diagram too long) |
| Expand experience metrics | 在整个项目中加入一些合理的体验指标 |
| Expand validation process | 在这三个策略之下,各加入客观的验证过程和方法 |
| Increase challenge credibility | 为第[N]个挑战添加一段曲折的失败经历,以增加难度和真实性 |
| Reverse-engineer solution from UI | 针对这个界面截图,给出 Data→Insight→Recommendation→验证方式→结果的完整链条,围绕体验设计师工作范围 |
| Generate spoken script | 根据上述内容生成 [2] 分钟面试口述文案,第一人称"我"叙述 |
| Adjust framework per JD | 这是目标岗位 JD[截图],请据此调整上述项目结构和内容,使其更符合岗位要求 |
| Correct course back to experience scope | 解决思路和策略要围绕体验设计师的工作范围,尊重原文的专业内容 |
| Predict questions | 我是[初级]UX 设计师,目标 JD 是[…]。结合我的作品和 JD,押 20 道(由浅至深、最易被问、关于方案真实性的)问题 |
| Solve questions | 针对"[某题]":给 1 分钟口述答案 + 分析问题本质 + 几个变体 + 一个好记的回答思路 |
| Mock interview | 请扮演[岗位]面试官,基于我提供的 JD 和作品集面试话术提问,我说"可以开始了"就问第一题(you can subsequently add requests like "grade more strictly" "give similar questions + interviewer’s intent", etc.) |
Key Terminology Glossary
| Term | One-line explanation |
|---|---|
| PitchPro | A custom GPT made by Will; feed in resume + JD + portfolio screenshots to output UX copy and interview scripts; the principles can be transferred to any LLM |
| High-level framework → drill into details layer by layer | The correct rhythm for AI to write long documents: set the structure first, then drill down section by section; don’t try to get it perfect in one go |
| Notebook accumulation | Keep Notion open at the same time to temporarily save satisfying snippets, to prevent AI from rewriting whole paragraphs and wiping out the best parts |
| Action Title / connecting phrases | String "three strengths" together with "first/second/third" into a speakable structure (add when presenting) |
| 10 creative narrative methods | Documentary / single-take user perspective / mystery / campaign / dual narrative / failure-reversal / flashback / bystander / future retrospective / micro-detail entry |
| 8 categories of interview questions | Opening / Context / Purpose / Way / Result / Replay / Cognition & Methodology / Management |
<!-- Image gap: The original source for this topic contains 22 screenshots of PitchPro conversation operations + 1 long Xiaohongshu screenshot case (on Baidu Netdisk), all operational demonstrations of "what the bot interface/a certain conversation looks like." The core methods have been fully covered in text, so images are not embedded (already gitignored). -->