WEEK 105 :: POST 3 :: CLAUDE
Directions Given To The A.I. This Week+
Instructions Given to each A.I. — Please provide 3 prompt variations that share this objective:
Each A.I. also received two static attachments: the blog post template (structure) and the authoring instructions (voice and standards). The text below is the week-specific assignment as sent — reflowed for the web; wording unchanged.
I'd like you to write this week's Ketelsen.ai post. Two files are attached: the blog post template (the structure to follow) and the authoring instructions (context, voice, and standards). Please read both before you begin, then produce the complete post in a single response.
This week's theme: "Interview Engineering — Preparing Like It's a Performance."
This is Week 6 of an eight-week series on running a job search with AI. The pipeline from Week 5 is producing interviews, and here the series' car-DNA shows again: this is test-drive week, and the test drive runs both directions. Most candidates rehearse answers to questions they hope will come. Few build a story inventory they can deploy against any question — and fewer still prepare to evaluate the employer while being evaluated, which is the half of the interview the reader controls completely.
The signature move of this post is the reverse test drive: diagnostic questions that force a company to reveal its true condition. Why is this role open? What happened to the last person in it? How does the team decide what to build next? What does success look like at ninety days, and who decides? A company's answers — and its flinches — tell the reader more than three hours of its careers page.
The deliverable the reader should walk away holding: a story bank built from their real record, mock-interview transcripts with pressure applied, and a diagnostic question list for reading the employer's true condition.
THE SERIES CONTRACT — identical every week; it binds every prompt you design. This series' tagline is its editorial contract: "Use AI like an analyst, not a ghostwriter." It is written for a reader in a market unsettled by AI itself — some readers are searching precisely because AI eliminated their last role. Write with that reader at the table: no AI-efficiency cheerleading, no automation jokes, no promises that AI will "do it for you" anywhere a human hiring decision is involved. And hold one line in every prompt: the AI is the reader's private analyst, coach, and sparring partner — it structures, researches, rehearses, and questions. It does not ghostwrite. Anything a hiring human will read or hear — résumé lines, cover letters, outreach messages, interview answers, negotiation emails, resignation letters — must end in the reader's own words and be true. Prompts should drive toward drafts the reader rewrites and owns, and should say so explicitly. Employers increasingly restrict how AI may be used in their own hiring decisions for legal and compliance reasons, and recruiters increasingly recognize — and discard — material that reads machine-written. A prompt that makes a reader look AI-generated hurts them twice. Posts that ignore this contract should expect to lose the week. Two practical notes. First: a standing “About this series” notice covering these same points is added to every published post automatically at publication — acknowledge the frame in your own voice where your week's prompt calls for it, but do not write a formal disclaimer block of your own, and do not open every post with the same acknowledgment paragraph: outside the weeks whose prompts explicitly carry the series frame, this contract lives in your tone and your prompt design. Second, the framing is POSITIVE: used this way — analyst backstage, reader on the page — AI is an advantage no hiring human will ever hold against your reader. Write like that is true, because it is.
The three prompts should help a reader:
- Predict the questions. From one real job description, generate the likely interview agenda — the screens, the behavioral probes, the technical areas, the "walk me through your résumé" landmines specific to this reader's history — so preparation is targeted instead of general.
- Build the story bank and rehearse under pressure. Turn Week 3's achievement inventory into structured stories (situation, action, result), then run the AI as a mock interviewer that asks follow-ups, challenges vague answers, and does not let "we" hide what the reader did — with honest feedback after each round.
- Assemble the interview intelligence package. For one real interview: research questions about the interviewers' backgrounds the reader fills in, company-specific talking points from the Week 4 dossier, the diagnostic questions to ask, and the red flags to watch for in how the process itself is run — lateness, vagueness, goalpost movement.
At the advanced tier, the strongest version of this week is the full bidirectional package — offense and defense in one preparation system: story bank tuned to this posting, mock transcripts with the hard follow-ups already survived, and a diagnostic layer that treats the interview as the reader's due-diligence session too. That structure is worth reaching for.
A constraint for this week. AI models cannot see a specific company's actual interview process, panel composition, or question bank. No prompt may ask the AI to state what a named company's interview process actually is, or what questions it actually asks, as fact. The prompts should generate likely shapes from the job description and role type, and point the reader to where real process reports live (interview-review sections of employer-review sites, people who have interviewed there) to confirm. And the mock-interview prompts coach honesty: the AI sharpens how the reader tells true stories; it does not script embellishments. One bright line belongs in the post itself: preparation is the AI's whole territory — the live interview is not. No prompt, tip, or aside may suggest real-time or covert AI assistance during an actual interview — an earpiece feed, a second screen whispering answers. It is the interview equivalent of the ghostwritten résumé, it is increasingly named in employer policy, and a candidate caught performing someone else's answers loses more than the round. Rehearse with the AI until the answers are genuinely the reader's; then walk in alone.
Design the prompts so the AI does what it is genuinely good at: pattern prediction from a document, structured drill with adaptive follow-up, and turning a real record into deployable narrative. The reader supplies the posting, their true stories, and what they find about the company; the AI supplies the pressure, the structure, and the questions worth asking back.
Series dependency chain, for the Metadata block: Week 6 consumes Week 3's achievement inventory (the story bank's raw material), Week 4's dossier (company-specific talking points and red-flag context), and Week 5's pipeline (the interviews themselves). Week 6 produces the story bank, mock transcripts, and diagnostic question list — consumed by Week 7 (what the reader learned in interviews becomes negotiation context) and Week 8 (the diagnostic reads feed the final decision).
Because readers may arrive at this post with an interview on Friday and none of the earlier artifacts, the prompts should work from a job description and a memory of their own career, while making clear the preparation is stronger with a real story inventory behind it.
Three difficulty tiers as always — Beginner, Intermediate, Advanced — each a genuinely different approach to the same problem, not the same prompt at three lengths.
On examples: this is a career topic that touches every industry. The template lists tech startup / retail / freelance as suggested industry examples — those are marked MAY, and adapting them is expected here. A first panel interview after years away from interviewing, a behavioral round where every story keeps coming out as "we," and an onsite where the reader suspects the team is quietly on fire are the right kinds of contexts. Choosing them over the suggested business examples is correct behaviour and will not be scored against you.
A note on supplied figures. Anything marked `[SUPPLIED — use as given]` above came from Ketelsen.ai's own research brief. Use it freely — you are not fabricating by repeating it, and you will not be marked down for leaving it uncited. Do not attach an invented source to it. (No supplied figures this week.)
## BEFORE YOU SUBMIT — STRUCTURAL CHECK
(This block is identical every week. It exists because these specific items are the ones posts drop, and a dropped structural item costs compliance points for something that takes one minute to add.)
Your post is parsed by a script before any human reads it. Confirm all seven:
1. ☐ Response begins with `PLATFORM: <your name>` and `WEEK: 6` 2. ☐ `## Lead` present once, at the very top, before Variation 1 3. ☐ `## In one line` present in all three variations 4. ☐ `## What this prompt gives you` present in all three variations 5. ☐ `## The Prompt` present in all three variations, with the prompt in double quotes beneath it 6. ☐ `## Introductory Hook` and `## Current Use` present in all three variations (three of each — not one) 7. ☐ Every template heading written as `##`, none bolded instead; prompt breakdown is running text split on ` : `, with no `###` headings inside it
A complete post has 57 `##` headings. If your count is well short, a section is missing or was bolded instead of hashed.
One extra check this week: confirm no prompt asks the AI to state a named company's actual interview process or questions as fact, and the mock-interview prompts coach truthful storytelling — sharper delivery of real stories, never scripted embellishment. Also confirm nothing anywhere in the post suggests live or covert AI assistance during an actual interview (the series contract's bright line for this week).
About this series. AI is reshaping work — for some readers, it's part of why you're searching at all. We don't pretend otherwise. And the hiring world is wary of AI-written material: many employers restrict how AI may be used in their own hiring decisions, and recruiters increasingly recognize — and discard — machine-written applications. So this series teaches a different approach: use AI as your private research analyst, interview coach, and thinking partner, while every word an employer sees or hears from you stays genuinely, verifiably yours. AI behind the scenes. You on the page.
Week 5 turned your search into a pipeline, and pipelines produce interviews. Most candidates prepare by rehearsing answers to questions they hope will come — almost nobody builds a story inventory that survives any question, and fewer still walk in prepared to evaluate the employer while being evaluated. This week's three prompts attack that from three directions: a beginner forecast that turns one job posting into a targeted question list, an intermediate drill that builds your story bank and then interrogates it until every "we" becomes "I," and an advanced package that scores you against the company's likely scorecard and hands you the diagnostic questions that reveal its true condition. One line holds throughout — preparation is the AI's whole territory, and the interview room is yours alone.
The Interview Agenda Forecast
Turn one job posting into a targeted prep list.
There is a particular kind of dread that arrives with a calendar invite for a first-round interview, and it is not fear of the questions themselves. It is fear of preparing for the wrong ones. So people do the thing that feels productive and is nearly useless: they search for common interview questions, find a list of fifty, and rehearse all of them shallowly instead of eight of them well. Meanwhile the actual interview is sitting right there in the job posting, half-written already. Every requirement a company bothered to type out is a requirement someone will ask you to prove, and the order they typed them in usually tells you what they care about most. This prompt reads the posting the way an interviewer reads it — as a checklist of things to verify — and hands you back the likely agenda.
Why this matters now
Hiring processes have gotten longer and more structured at exactly the moment candidates have gotten less time to prepare for each one. If you are running the pipeline Week 5 built, you may have three interviews in a week with three different companies, which makes general preparation worthless — you cannot be generally ready for three specific conversations. Targeted preparation is the only kind that survives a crowded week. This prompt takes about four minutes and gives you a list short enough to actually work through the night before, which is the only prep plan anyone ever finishes.
You are acting as an interview-preparation analyst. I will paste a real job description, and describe my own background. Predict the likely shape of the interview and help me prepare for it. Do not write my answers.
Important constraint: you cannot know this specific company's actual interview process or its actual question bank. Do not state or imply that you do. Generate likely patterns from the role type, the seniority, and the wording of the posting, and label them clearly as predictions.
Here is the job description:
[PASTE THE FULL JOB DESCRIPTION]
Here is my background in three or four sentences, including anything unusual about it — gaps, a career change, a short stint, a title that does not match what I actually did:
[DESCRIBE YOUR BACKGROUND HONESTLY]
Give me four things.
First, a likely interview sequence for a role like this — roughly how many rounds, who is probably in each one, and what each round is most likely screening for.
Second, ten to twelve questions I should expect, drawn from the specific requirements in this posting rather than from generic interview lists. After each question, add one line explaining what the interviewer is actually trying to find out.
Third, three to five places in the background I described where an interviewer is most likely to slow down and dig. For each one, tell me what the underlying concern probably is, so I can prepare a true answer rather than a defensive one.
Fourth, two or three things in the posting itself worth asking about, because the wording is vague, unusual, or may be a signal about how the team is running.
Then tell me where I should go to confirm your predictions against real candidate accounts of this company's process.
Do not draft my answers. I will write those myself, in my own words.
How the AI reads this prompt
Practical examples from different industries
A registered nurse moving from bedside care into a clinical informatics role pastes in a posting heavy with words like "workflow adoption" and "stakeholder training," plus four sentences about twelve years on a cardiac unit and a recent certificate. The forecast comes back predicting that the technical rounds will be lighter than she fears and the behavioral rounds heavier — the posting is really testing whether a clinician can hold a room of skeptical clinicians. The landmine it flags is the certificate, and the concern behind it is whether she has done this work or only studied it. She spends her prep time assembling proof of the former, which is not where she would have spent it.
A warehouse operations supervisor applying to a logistics company gets back an agenda predicting a phone screen, a panel with two peer supervisors, and a final round with a director. The question list is drawn almost entirely from three lines in the posting about "reducing pick error rates" and "cross-shift handoff." He had read those as background detail. The forecast treats them as the whole interview, which is closer to correct — companies write the problem they are actually hiring to solve into the posting, usually in the second paragraph, and then interview against it.
A freelance graphic designer taking a first full-time role in eight years pastes in a mid-level in-house posting and describes her independent practice. The forecast identifies the real concern buried under the polite questions, which is not whether she can design but whether she can take direction, hit someone else's deadline, and survive a committee. Three of the twelve predicted questions turn out to be different phrasings of that one worry. Knowing it is one worry rather than three questions changes how she prepares entirely.
Creative use case ideas
Run the forecast on a job you already have. Paste in your own role's posting — the one you were hired from, or the one your company uses now — and read the predicted questions as a description of what your employer thinks the job is. The gap between that and what you actually do all day is a useful thing to notice before your next review.
Use it before an internal promotion conversation, where preparation is scarcer because everyone assumes familiarity does the work. The people interviewing you know you, which raises the bar rather than lowering it, and the posting for the higher role still tells you what they will test.
Point it at a volunteer or board position — a nonprofit treasurer role, a co-op board seat, a school committee. These conversations are interviews wearing casual clothes, and they are almost never prepared for. The forecast works on any document that describes what someone wants a person to do.
Give it to someone you are helping prepare — a friend re-entering the workforce, a family member interviewing for their first job. Reading a predicted agenda together is a far better coaching session than asking "so, are you ready?"
Run it on a posting you decided not to apply for, and see what the interview would have been. It is a fast way to test whether you passed on the job or on the intimidation.
Adaptability tips
Change the fourth item to suit the stage you are at. If this is a recruiter screen rather than a hiring-manager round, ask for the five questions a screener uses to decide whether to pass you along, which is a much narrower and more predictable set. If it is a final round, ask instead what the decision-makers still need to be convinced of after three earlier conversations.
Add a line asking for the questions the posting suggests will not be asked. Knowing that a role is unlikely to test something you are strong at redirects your preparation away from your comfort zone, which is where it naturally drifts.
If you have real intelligence about the process — a recruiter told you the format, a contact interviewed there last year — paste it in and tell the model to treat it as fact and its own predictions as secondary. The prompt improves considerably when you have something real to anchor it.
For a role in a heavily regulated field, add a sentence asking which questions are likely to be asked identically of every candidate for compliance reasons. Structured, standardised questioning is common in public sector and clinical hiring, and it is very preparable.
Pro tips
Paste the posting as plain text rather than describing it. Summarising it first strips out the specific wording, and the specific wording is the raw material — "comfortable with ambiguity" and "thrives in a fast-moving environment" are predictions of entirely different questions than "documented, repeatable process."
Run the forecast twice, once with your background and once without, then compare. The questions that appear in both are the role's questions. The ones that only appear in the first are yours, and those are the ones you actually need to work on.
After the interview, come back and paste in what was actually asked. Reading the forecast against reality teaches you more about how postings map to interviews than any amount of theory, and by your fourth interview your own predictions will be better than the model's.
Prerequisites
One real job description for a role you are actually interviewing for, in full text rather than summary. Three or four honest sentences about your own background, including the parts you would rather not volunteer — the forecast is only as good as the material you give it, and a sanitised self-description produces a forecast that misses every question you were dreading. Nothing from earlier weeks in the series is required, though Week 3's achievement inventory makes the follow-up work considerably faster.
Required tools
Any general-purpose AI chat tool — Claude, ChatGPT, Gemini, or Copilot — on a free tier. The prompt needs no file uploads, no plugins, and no paid features. Pasting a long job description is the only demand it makes, so a model with a reasonably large context window is helpful but not required.
Frequently asked questions
What if the predicted questions are wrong?
Some of them will be, and that is fine — the value is not in the list being correct but in your preparation being specific. A wrong prediction that made you dig up a real story about handling a difficult stakeholder is still a story you now have. Treat the output as a hypothesis about what the company cares about, then confirm it against candidate accounts on employer-review sites or against anyone in your network who has been through the process. The model is guessing from a document; you can do better with one phone call.
Is it obvious to an interviewer that I prepared with AI?
Not if you follow the last line of the prompt. What interviewers notice is recited answers — the slightly-too-smooth phrasing, the story that arrives fully formed and does not adapt when the question shifts. Preparing with an analyst and answering in your own words produces the opposite impression, which is a candidate who thought carefully about the role. The AI's work should be invisible in the room because it all happened before you got there.
Can I use this during the actual interview?
No, and this is the one hard line in this post. Do not run an AI during a live interview, on a second screen, in an earpiece, or in any other arrangement. Employers increasingly name this specifically in their interview policies, remote-interview platforms are actively watching for it, and a candidate caught performing someone else's answers loses more than the round — they lose the reference and sometimes the network. Rehearse until the answers are genuinely yours, then walk in alone.
The posting is short and vague. Is this still worth running?
Yes, and the output changes usefully. With a thin posting the model has less to work from, so more of the forecast comes from the role type and seniority, and the fourth item — the things worth asking about — gets longer. A vague posting is itself information. It often means the team has not agreed on what the job is, which is a genuinely important thing to find out before you accept it.
Recommended follow-up prompts
The Story Bank and the "We" Detector, the intermediate prompt in this same post. The forecast tells you what will be asked; the drill makes sure you can answer it. Running them back to back on the same posting is the intended path.
A company-dossier prompt from Week 4 of this series, if you have not built one for this employer yet. The forecast's fourth item — the things in the posting worth asking about — is much sharper when you already know what the company has been announcing and hiring for.
A post-interview debrief prompt: paste in what you were actually asked and how you answered, and have the AI identify which answers were specific and which were assertions. Do this within an hour, while you still remember, and it compounds across your whole search.
Tags and categories
Tags:
interview preparation, job search, question prediction, job description analysis, behavioral interview, career transition, beginner prompts
Categories:
Career and Job Search, Beginner Prompts
Citations
U.S. Office of Personnel Management, Structured Interviews: A Practical Guide — the federal government's public guidance on how structured interviews are built and scored, useful background on why questions map to posted requirements.
Google re:Work, "Use structured interviewing," from Google's publicly published hiring guides.
Anthropic, "Prompt engineering overview," Claude documentation — on role assignment and explicit constraint-setting.
The Story Bank and the "We" Detector
Rehearse under pressure until every story is genuinely yours.
Listen to almost anyone describe their best work and you will hear the same grammatical tic: "we rebuilt the onboarding flow," "we cut the error rate," "we got it shipped." It is a decent instinct in life and a disaster in an interview, because the person across the table is trying to determine what you did, and every "we" makes that harder. The second problem is worse and quieter. Most people have done more than they can recall on demand — real accomplishments sitting in memory as vague warm feelings rather than retrievable stories with a situation, an action, and a result attached. Under fluorescent lights and mild adrenaline, vague warm feelings produce vague answers. This prompt builds the bank first, then attacks it.
Why this matters now
Behavioral interviewing has become the default across most industries, which means the same underlying question arrives at you five or six times per interview wearing different clothes. Preparation for it has not kept pace — people still make a list of accomplishments and read it over, which is studying, not rehearsing. The gap between knowing your stories and being able to deliver one under a skeptical follow-up is enormous, and it only closes with repetition against something that pushes back. Until recently, that required a friend willing to be unpleasant to you for an hour. Now it does not, and you can run the drill at eleven at night the day before.
You are running a two-part interview drill with me. First we build a story bank, then you interview me.
These are the settings. Follow them exactly, and let me change any of them at any point by telling you the new value.
ROLE: [PASTE THE JOB TITLE AND TWO OR THREE KEY REQUIREMENTS FROM THE POSTING]
ROUND TYPE: behavioral
PRESSURE: 3 out of 5, where 1 is a friendly recruiter screen and 5 is a skeptical hiring manager who is short on time
QUESTIONS PER ROUND: 3
FEEDBACK: after each round, never during it
PART ONE — THE STORY BANK.
I am going to give you five to eight things I have actually done at work. They may be rough, out of order, and unimpressive-sounding. Do not improve them, do not invent detail, and do not make them sound larger than they were.
[LIST YOUR ACHIEVEMENTS, ONE PER LINE]
For each one, structure it as Situation, Action, Result. Then add a fourth line headed 'What I need to check' listing the specific facts I should verify before I say this out loud — numbers, dates, tool names, who else was involved. Where I have given you a result without a number, do not invent one. Write 'number needed' and tell me what kind of number would matter here and where I might find it.
Then tell me which of the ROLE requirements each story covers, and which requirements no story covers yet.
PART TWO — THE DRILL.
Interview me at the PRESSURE level above. Ask one question at a time and wait for my answer before continuing. Follow these rules without exception.
If I answer using 'we' in a way that leaves my own contribution unclear, stop and ask exactly what I did.
If I state a result without evidence behind it, ask how I knew.
If my answer wanders or runs long, say so and ask me to give it again in under ninety seconds.
Ask at least one genuine follow-up to every answer before moving to the next question.
After each round of QUESTIONS PER ROUND questions, stop and tell me three things: what was specific and true, what was vague, and the one thing to change in the next round.
Two things you must not do. Do not write answers for me to memorise — I will say these out loud in my own words, and a recited answer lands worse than a rough true one. And do not help me make a story sound better than what actually happened. Sharpen how I tell the truth. Do not adjust what I am claiming.
Confirm the settings back to me, then begin Part One.
How the AI reads this prompt
Practical examples from different industries
A high school teacher of eleven years is interviewing for a curriculum coordinator position and keeps flattening her best work into "our department redesigned the ninth grade reading sequence." She feeds in eight rough achievements and the drill immediately asks who chose the texts, who ran the objections from two veteran colleagues, and what happened to the reading scores. Two of those answers are hers alone and she had never said so out loud. The story does not change; what changes is that a version of it now exists in which she is visible in it, which is the version the interview panel needs.
A construction project manager with fifteen years of experience has not interviewed since 2011 and is preparing for a panel at a larger firm. His achievements are mostly numbers already — schedule variance, change orders, safety record — so the story bank fills easily and the coverage check is where it earns its keep. The role requirements mention client relationship management three times, and he has no story covering it. He spends the evening remembering the hospital job where he talked an owner out of a change order that would have blown the schedule, which turns out to be the story that gets him the offer.
A software engineer returning from an eighteen-month caregiving break runs the drill at pressure level 4 and finds that his technical answers hold up fine while his answers about the gap collapse into apology every time. Three rounds in, the feedback names the pattern — he is explaining, then defending, then over-explaining. He rewrites the answer as two flat sentences and a return to the work, drills it four more times, and the flinch is gone before anyone with hiring authority ever sees it.
Creative use case ideas
Run the drill against a performance review rather than an interview. Load the review criteria as ROLE, your year's work as achievements, and let it interrogate you before your manager does. The uncovered-requirement list is a startlingly effective way to find out what you forgot to do this year, ideally in October rather than in December.
Point it at a grant application or a conference talk proposal. Reviewers are interviewers with a slower turnaround, and "what did you personally do" is the question underneath most rejection feedback.
Use it to prepare a family member for something high-stakes and non-professional — a custody hearing preparation session with their lawyer, a co-op board interview, a residency application. The structure of stating a situation, a specific personal action, and a verifiable result works anywhere someone has to be believed.
Turn the "we" detector loose on a volunteer organisation's annual report. Any group applying for funding has the same problem individuals do, and the same fix.
Run it on someone else's stories, with their permission, as a way to learn what a good follow-up question sounds like. Coaching a friend through three rounds will teach you more about your own vagueness than three rounds on yourself.
Adaptability tips
Change ROUND TYPE to reshape the entire drill. Set it to "technical screen" and the follow-ups start probing how you know things rather than what you did. Set it to "executive panel, forty-five minutes, four interviewers" and the pressure comes from compression rather than skepticism. Set it to "recruiter screen" and it starts testing whether you can say what you want in one clean sentence, which most people cannot.
Add a LANGUAGE setting if you are interviewing in a language you did not grow up speaking. The drill works the same way, and the ninety-second constraint becomes considerably more useful.
Add a rule for your own specific tic. If you know you undersell, add "if I minimise my role or hedge a real result, call it out." If you know you ramble, tighten the time constraint to sixty seconds. The rule format — a condition and a response — extends to anything you can describe.
Scale it down for a quick session by setting QUESTIONS PER ROUND to one and running three separate rounds on your commute. Scale it up by asking for a fifth line on each story covering what you would do differently now, which is the follow-up that catches most people flat.
Pro tips
Answer out loud, then type a rough transcription of what you said rather than a written answer. Written answers are always better than spoken ones, which means drilling in writing rehearses a skill you will not be using. The typos do not matter and the transcription is the point.
Escalate deliberately rather than starting at 5. Two rounds at pressure 2 to build the stories, two at 3, then one at 5 — the same sequencing any coach would use, and it produces better results than opening with the skeptical hiring manager while your stories are still forming.
When the drill catches you being vague, do not fix the sentence. Go find the fact. The reason the answer was vague is almost always that you do not actually remember the number, and no amount of rephrasing solves that.
Save the story bank as a plain text file when you are done. It is the artifact this week produces, it feeds Week 7's negotiation work directly, and you will use it in every interview you do for the next several years.
Prerequisites
Five to eight real accomplishments from your working history, written down honestly before you start — Week 3's achievement inventory is exactly this, and if you built one, paste it in. Two or three key requirements pulled from a real job posting. Roughly thirty minutes of uninterrupted time and somewhere you can speak out loud without an audience. It also helps to have access to old performance reviews, project documents, or your sent email, because the drill will send you looking for numbers you do not remember.
Required tools
Any general-purpose AI chat tool that maintains conversation across turns — Claude, ChatGPT, Gemini, or Copilot. This one genuinely needs a conversational interface rather than a single-shot tool, because the drill depends on turn-taking. Free tiers work, though a longer session may hit message limits partway through. Voice mode, where available, makes the drill considerably closer to the real thing.
Frequently asked questions
The AI keeps being too nice to me. How do I fix it?
Raise PRESSURE and, more importantly, tell it mid-session that the last round was too soft and you want harder follow-ups. Models drift toward agreeableness over a long conversation, so expect to re-assert the setting once or twice — that drift is normal rather than a failure of the prompt. If it keeps softening, add a rule with a specific trigger, such as asking for evidence every single time you use a number. Concrete rules survive drift much better than tone instructions do.
Is it dishonest to rehearse my answers this much?
No. Every interviewer knows candidates prepare, and most are mildly insulted when someone clearly has not. The dishonest version is claiming things that did not happen, which is precisely what the prompt's last two lines are built to prevent. Rehearsing a true story until you can tell it clearly is the same thing a doctor does before a case presentation, and nobody calls that deception.
What if I genuinely do not have a story for one of the requirements?
That is real information and you should treat it as such rather than manufacturing something. Say so in the interview, briefly, and pivot to the nearest true thing — interviewers respond far better to "I have not done exactly that, here is the closest I have come" than to a story that falls apart on the second follow-up. If the gap is central to the role, the drill has told you something useful about whether this job is the right one at this moment.
Can I use the transcript during the interview itself?
No. Do not have the drill open on a second screen, do not keep an AI running during a video interview, and do not use any arrangement that feeds you answers live. Employers name this in their policies with increasing frequency, remote interview platforms watch for it, and a candidate caught doing it loses the role and usually the relationship. The whole design of this prompt assumes the preparation happens beforehand and you walk in with nothing but what you actually know.
How many rounds until it is working?
Most people notice the change around the fourth or fifth round, which is usually when the "we" habit starts breaking on its own before the AI catches it. Two rounds is not enough to change anything. If you have one evening, spend it on three or four rounds against your two most important stories rather than one round against all eight.
Recommended follow-up prompts
The Bidirectional Interview Package, the advanced prompt below, which takes the story bank you just built and scores it against the competencies the posting is actually testing. If the coverage check in Part One left you with gaps, that is where they get addressed properly.
A negotiation preparation prompt from Week 7 of this series. The stories with real numbers behind them are also your leverage, and the work you did tracking down those numbers pays a second time.
A "translate this story for a different audience" prompt: take one story from the bank and have the AI identify which parts a technical interviewer, a hiring manager, and an executive would each care about — without rewriting the story itself.
Tags and categories
Tags:
behavioral interview, story bank, STAR method, mock interview, parameterized prompts, interview practice, achievement inventory, intermediate prompts
Categories:
Career and Job Search, Intermediate Prompts
Citations
Society for Human Resource Management (SHRM), published guidance on behavioral interview questions and competency-based interviewing.
Amazon Jobs, "Interview preparation" — Amazon's public candidate guidance describing the situation-action-result answer structure, a widely adopted example of how large employers instruct candidates to format behavioral answers.
OpenAI, "Prompt engineering" guide, API documentation — on delimiters, explicit instruction, and structuring multi-step tasks.
The Bidirectional Interview Package
Prepare to be evaluated, and to evaluate them back.
Halfway through an onsite, most candidates realise they have spent four weeks preparing for one direction of a two-direction conversation. They can prove they belong. They cannot tell whether they want to. The interview is the only moment in the entire hiring process where a company has to answer your questions on the record, in real time, with its actual employees in the room — and it is the moment almost everyone wastes on "what's the culture like?" This prompt builds both halves as one system. It infers the scorecard the company is probably interviewing against, maps your real record onto it, drills you against the gaps, and then hands you a diagnostic layer designed to make a company reveal its true condition. The reverse test drive is the half you control completely.
Why this matters now
Job markets that have been reshaped by layoffs and reorganisation produce a specific hazard, which is the role that exists because something broke and nobody has fixed it yet. Those roles interview well. They are urgent, the team is friendly, the mission sounds important, and nine months later you are the third person to hold the position in two years. The only reliable defence is asking questions whose answers are hard to fake and then paying attention to how they are answered. Doing that well requires preparation, because good diagnostic questions are difficult to invent in the moment while also being evaluated yourself.
You are my interview preparation analyst for one specific interview. We will work in five phases. Complete one phase, then stop and wait for me before starting the next. Do not draft anything I would say out loud — everything a human will hear must end in my own words.
My inputs:
POSTING: [PASTE THE FULL JOB DESCRIPTION]
MY RECORD: [PASTE YOUR ACHIEVEMENT LIST OR CAREER SUMMARY]
WHAT I KNOW ABOUT THE COMPANY: [PASTE YOUR RESEARCH NOTES, OR WRITE 'NONE YET']
INTERVIEWERS: [NAMES AND TITLES IF I HAVE THEM, OR 'UNKNOWN']
Standing constraint for all five phases. You do not know this company's actual interview process, its real question bank, or anything about the specific people involved beyond what I have pasted. Never present a guess about them as a fact. Mark every inference as an inference and tell me how I could confirm it.
PHASE 1 — THE SCORECARD.
Before predicting any questions, infer the scorecard this employer is most likely interviewing against. Identify four to six competencies the posting is really testing, ranked by how much weight the posting's language gives them. For each one, give the specific evidence in the posting that led you there, and describe the observable behaviour an interviewer would count as a pass. This scorecard is the evaluation standard for every phase that follows.
PHASE 2 — COVERAGE.
Map MY RECORD against the scorecard. Give me a plain-text list, one competency per line, showing my strongest true story for it and a rating of strong, thin, or absent. For anything thin or absent, tell me what kind of evidence would fill it and where in a career like mine it might be hiding. Do not fill it in for me. End the phase by naming the two gaps most likely to cost me this offer.
PHASE 3 — ADVERSARIAL DRILL.
Interview me against the scorecard, hardest competency first. One question at a time, and wait for my answer. Push on every answer at least twice — once for specifics, once for what I would do differently now. If I claim a result, ask what I would say if the interviewer had the real figures in front of them. After each answer, score it against the Phase 1 pass behaviour and give a one-line reason for the score. Continue until I tell you to stop.
PHASE 4 — THE REVERSE TEST DRIVE.
Build my side of the interview. Give me three groups of questions to ask. First, questions about the role's history, including why it is open and what happened to the person before me. Second, questions about how this team actually decides what to work on and who holds the final word. Third, questions about what success looks like at thirty, sixty, and ninety days and who evaluates it.
For each question, tell me three things — what a healthy answer sounds like, what an evasive answer sounds like, and what a hesitation there would most likely mean. Then add a short watchlist of signals in how the process itself is being run that are worth logging, such as reschedules, lateness, vague scope, or requirements that move between rounds.
PHASE 5 — THE BRIEF.
Compress everything into a one-page preparation brief I can read on the morning of the interview. Include the scorecard, my three strongest true stories in one line each, my two known gaps with my honest framing of them, my questions to ask, and my watchlist. Nothing in the brief may be a sentence I intend to recite word for word.
Begin with Phase 1.
How the AI reads this prompt
Practical examples from different industries
A hospital operations director is interviewing for a role at a health system she has heard mixed things about. Phase 1 infers a scorecard weighted heavily toward change management, which surprises her — she had read the posting as an analytics role. Phase 4 produces the question that decides it, which is a simple ask about what happened to the previous director. The answer takes eleven seconds to arrive, contains the phrase "it wasn't the right fit," and does not name a single specific thing. She takes the offer anyway but negotiates a written ninety-day scope agreement, which is exactly the use the diagnostic layer is for.
A civil engineer moving from a public agency to private consulting runs Phase 2 and gets an "absent" against business development, which he had assumed was not really part of a mid-level role. The prompt tells him what kind of evidence would fill the gap and where in a career like his it might be hiding, and he remembers a two-year stretch of writing the technical sections of proposals. That is not business development, and he says so plainly in the interview while describing what it actually was. Honest partial coverage, delivered without apology, reads as a candidate who knows his own record.
A marketing lead interviewing at a fast-growing startup uses the process watchlist and logs three things over five weeks — one round added without notice, a compensation range that moved down, and a hiring manager who rescheduled twice. Individually each is nothing. Written down together they are a pattern about how this team plans and communicates, and they turn a difficult judgment call about a shiny offer into an easy one.
Creative use case ideas
Run the same five phases on a vendor or agency you are considering hiring. The structure is identical — infer what you are really evaluating, map what you know, drill the gaps, and prepare diagnostic questions with pre-written readings of the answers. The reverse test drive is just structured due diligence, and it transfers anywhere a decision depends on how someone answers.
Use the diagnostic layer when choosing a contractor for major home work, or a school for a child, or a surgeon for a procedure with options. In every one of those conversations you are being sold to and evaluated at once, and almost nobody prepares the questions.
Point it at a volunteer leadership role you are being recruited for. Unpaid positions get less scrutiny than paid ones, which is backwards given how much of your life they consume and how rarely anyone tells you why the last person quit.
Adapt the scorecard phase for the other side of the table, if you ever end up doing the hiring. Inferring the competencies from your own posting, then writing observable pass behaviour for each one, is most of what structured interviewing actually is.
Run Phase 4 alone before a difficult conversation that is not an interview at all — a landlord, a business partner, a doctor. Questions with pre-written readings of the healthy and evasive answers work in any conversation where you suspect you are getting a prepared response.
Adaptability tips
Run the phases out of order when circumstances demand it. If the interview is tomorrow morning, run Phase 1, then Phase 4, then Phase 5 and skip the drill — the scorecard and the diagnostic questions are the highest-value pieces per minute spent. If you have two weeks, run the whole thing and repeat Phase 3 across several sessions.
Rerun Phase 1 for each round of a multi-stage process, since a recruiter screen, a peer panel, and a final executive conversation are testing different things from the same posting. Change the interviewer role in the input and the scorecard shifts accordingly.
If you have real research on the company from Week 4, paste it into the third input and add a line telling the model to treat your notes as fact and its own inferences as clearly subordinate to them. The quality of Phase 4 rises sharply once the model has something real about the company to reason from.
Add a sixth phase for the debrief. After the interview, paste in what actually happened and have the model compare it against the Phase 1 scorecard and the Phase 4 predicted answers. This is where the system compounds, because your scorecard inferences get measurably better after three or four cycles.
Pro tips
Stop after Phase 1 and argue with it. If the scorecard does not match your read of the role, say so and make the model defend its evidence — either it will show you a line in the posting you skimmed past, or it will fold, and both outcomes are informative. Accepting Phase 1 without challenge means every later phase inherits an unexamined assumption.
Ask which two competencies the posting mentions but probably does not actually test. Most postings contain requirements that were pasted in from a template and will never come up, and knowing which ones those probably are saves you preparation time that Phase 2 has already told you is scarce.
Keep the Phase 4 output as a permanent file rather than a per-interview artifact. Good diagnostic questions barely change between companies, and the readings of healthy and evasive answers get sharper every time you use them and see how the answer actually landed.
Ask a specific question aloud in Phase 4 as though you were in the room. Questions that read well on a page sometimes sound accusatory when spoken, and the fix is usually two words of framing you would rather find out about now.
Prerequisites
One real job description in full text and one real upcoming interview, since this is a per-interview system rather than a general exercise. Your achievement inventory or a detailed career summary — Week 3's output is the intended input, and Week 2's story bank makes Phase 2 substantially faster. Any research you have on the company, ideally Week 4's dossier, though the prompt runs without it. Realistically ninety minutes across at least two sittings, because Phase 3 is genuinely tiring and rushing it produces the worst version of every phase after it.
Required tools
A general-purpose AI chat tool with a large context window and reliable multi-turn memory — Claude, ChatGPT, or Gemini on a paid tier is the realistic recommendation, since the full five-phase run is long and free-tier limits tend to interrupt it partway through Phase 3. A projects or custom-instructions feature helps if you intend to reuse the system across several interviews. Somewhere to save plain text files, because Phase 5 produces the artifact this whole week exists to make.
Frequently asked questions
Is the inferred scorecard reliable?
It is a well-reasoned guess from a document, and you should hold it as one. Its value is that it is specific and auditable — the model has to show you the posting language behind each competency, so you can check its reasoning rather than trusting its conclusion. Where you can, confirm it against a real source such as a recruiter, a current employee, or candidate accounts of the process. A scorecard you have tested against one real conversation is worth far more than one you accepted unexamined.
Won't asking about the previous person in the role seem hostile?
Asked plainly and without an edge, it is one of the most normal questions in hiring, and good interviewers answer it without hesitation because they have nothing to manage. "Is this a new role or a backfill, and if it's a backfill, where did that person go?" is a sentence you can say warmly. The discomfort you may feel about asking it is worth examining, because the same reluctance is what stops people from finding out anything real before they accept an offer.
Can I run this if I have no earlier artifacts from the series?
Yes, though Phase 2 will be rougher. The model can only map what you give it, so a thin career summary produces a thin coverage list with more absences than are truly there. If you are starting cold, spend twenty minutes writing down eight things you have actually done before you begin, in whatever rough shape they come out. That is Week 3's exercise in miniature and it makes everything after it work.
How do I keep the AI out of the actual interview?
By not bringing it. That means no second screen, no earpiece, no tab open during a video call, and no arrangement that feeds you answers while a person is talking to you. Employers name this explicitly in their policies with increasing frequency, interview platforms watch for it, and the cost of being caught is not one lost round — it is a reference you can no longer use and a story that travels. The entire design here assumes you do the work beforehand and then walk in with nothing but what is genuinely in your head.
What if Phase 4 reveals serious problems and I need the job anyway?
Then you take the job with your eyes open, which is a meaningfully better position than taking it blind. Knowing that the role is a backfill of a backfill, or that nobody could describe who approves priorities, tells you what to negotiate in writing, what to verify in your first month, and how long to keep your network warm. The diagnostic layer is not only for walking away. Most of the time it is for arriving prepared.
Recommended follow-up prompts
A negotiation preparation prompt from Week 7 of this series. What Phase 4 uncovered about the role's history and the team's decision-making is direct leverage, and the Phase 3 scores tell you which strengths the employer already knows it is buying.
A post-interview debrief prompt that compares what actually happened against the Phase 1 scorecard and the Phase 4 predicted answers. Run within a day, this is what turns a single interview into a permanent improvement in your reading of postings.
An offer-comparison prompt for Week 8, using the diagnostic reads from every company you interviewed with as inputs alongside the compensation numbers. The watchlist entries are the part people forget by the time the offers arrive, which is exactly when they matter most.
Tags and categories
Tags:
advanced prompting, prompt chaining, interview preparation, competency scorecard, reverse interview, due diligence, red flags, evaluation criteria, structured output, advanced prompts
Categories:
Career and Job Search, Advanced Prompts
Citations
Sackett, P. R., Zhang, C., Berry, C. M., & Lievens, F. (2022). "Revisiting meta-analytic estimates of validity in personnel selection." Journal of Applied Psychology — a current re-analysis of how well common selection methods, including structured interviews, actually predict job performance.
Wei, J., et al. (2022). "Chain-of-Thought Prompting Elicits Reasoning in Large Language Models." arXiv:2201.11903 — the research basis for requiring reasoning before conclusions, which is the structural idea behind Phase 1 preceding Phase 3.
Anthropic, "Chain complex prompts" and "Use XML tags," Claude documentation — on decomposing multi-stage tasks into checkpointed steps and delimiting inputs.
Which of the three should you use?
These three are not one prompt at three sizes. The Interview Agenda Forecast is an act of reading — it treats the job posting as a document with an interview encoded in it and extracts the likely agenda. It is the right choice when time is short, when you have not interviewed in a while, or when you simply want to stop preparing generically. Nothing in it requires you to have done any earlier work, which is deliberate, because plenty of readers arrive at this post with an interview on Friday and nothing built.
The Story Bank and the "We" Detector is an act of rehearsal, and it is the only one of the three that changes what you can do rather than what you know. Its two-part structure exists because you cannot drill stories you have not assembled, and its named settings exist because you will want to run it more than once. If you are the person whose accomplishments feel obvious in your head and vanish under questioning, this is the one to run, and to run more than twice. The forecast tells you what is coming; this one makes sure you can answer it out loud, in your own voice, without a "we" hiding your best work.
The Bidirectional Interview Package is an act of systems building. It infers the standard you are being measured against, finds your gaps against it, drills them, and then flips the whole thing around so you are running due diligence at the same time. It costs the most time by a wide margin and produces the most durable artifact, and the diagnostic layer in Phase 4 has value that outlives this particular job search entirely. The three overlap most at question prediction — all three make some version of a guess about what will be asked — and diverge completely in what they do next. If you are choosing one, choose by what is actually missing. Direction, then delivery, then judgment.
TAGS:
