What Your Daily Work Is Actually Worth on the Open Market
WEEK 100 :: 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: "Should I Even Be Job Searching Right Now?"
This is Week 1 of a new eight-week series on running a job search with AI — the follow-up to "AI at the Dealership," and it shares that series' backbone: a high-stakes, emotionally loaded decision where the other side of the table (employers, recruiters, applicant tracking systems) holds most of the cards. Week 1 starts where honest job searches start: before the résumé. Most people begin applying before they have defined what is actually broken about their current situation, and impulsive moves are how people land in roles they regret within months. The reader's job this week is to decide — stay, grow in place, or go — with evidence instead of a mood.
One thing this week must do that no later week has to: speak to all three readers who arrive at a job-search series. The active-but-employed reader wondering if the grass is greener, the reader who has just been laid off and has no stay option, and the career pivoter for whom "go" means a different field entirely. And name the elephant with care: for some laid-off readers, AI itself is part of the story — restructurings framed around AI efficiency are part of this market. A series about using AI must not be breezy about that; one honest, warm sentence acknowledging it (no layoff statistics — the constraint below already forbids them) buys more trust than a page of enthusiasm, and the laid-off reader should feel seen, not lectured.
Week 1 also opens the series, so its Lead carries the series' editorial frame. The Lead should say plainly what this series is and is not: AI as the reader's private analyst, coach, and thinking partner — never their ghostwriter — because hiring decisions are made by humans who are rightly wary of machine-written material, and employers themselves face legal and compliance limits on AI in hiring. The two lines the series stands on — "Use AI like an analyst, not a ghostwriter" and "AI behind the scenes. You on the page." — belong in or near this Lead, worked into the post's own voice rather than dropped in as slogans.
The prompts should handle the three audiences through a "MY SITUATION" context block the reader fills in — the same pattern the car series used — so one prompt serves all three without pretending they are the same person.
The deliverable the reader should walk away holding: a written stay-or-go decision, a personal compensation baseline (what they earn now, fully loaded), and a timeline — the three artifacts every later week in this series builds on.
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:
- Run an honest career audit. A satisfaction diagnostic that separates "bad month" from "bad fit" — role, manager, growth, compensation, energy — and names what specifically is broken, so the reader is diagnosing before prescribing.
- Inventory their skills and market position. What the reader actually does all day, translated into the language the market hires for, with an honest read on which skills are appreciating and which are aging — the AI structuring the inventory from the reader's own history, not asserting market statistics.
- Build the full stay-vs-go decision framework. The "budget week" of the series: current compensation fully decoded (base, bonus, benefits, the things that quietly vanish on exit), savings runway if the search goes long, the benefits cliff of leaving mid-year, weighed inside a decision framework the reader owns — ending in a written decision and timeline.
At the advanced tier, the strongest version of this week is a decision memo the reader writes to themselves — situation, evidence, options scored, decision, revisit date — produced by a prompt that makes the AI a structured interviewer and analyst rather than an oracle. A decision the reader can re-read in six months beats a vibe either way.
A hard constraint, stated up front for the series. AI models cannot see live labor-market data, current layoff patterns, or real salary postings, and this series' citation standards treat unverified statistics as defects. No prompt may ask the AI to state current hiring trends, layoff figures, salary levels, or "shift shock" regret statistics as fact. Where market reality matters, the prompt should have the reader supply what they know or point them to named sources to check — pay-transparency postings, official labor statistics, their own industry contacts — not have the AI assert numbers. The same applies to the money: the runway analysis works on numbers the reader supplies, and nothing in these prompts is financial advice — the framework organizes the reader's own decision, it does not hand down a verdict.
Design the prompts so the AI does what it is genuinely good at: structured interviewing, translating a work history into market language, organizing a messy emotional decision into evidence and options. The reader supplies their situation and their numbers; the AI supplies structure, candor, and sequence. Posts whose prompts have the AI invent market statistics or deliver quit-your-job verdicts should expect to be marked down on Practical Utility and Content Accuracy.
Series dependency chain, for the Metadata block: Week 1 consumes nothing — it is the series opener. Week 1 produces the written stay-or-go decision, the compensation baseline, and the search timeline — consumed by Week 2 (defining the target), Week 7 (the baseline anchors the negotiation), and Week 8 (the decision criteria return in the final offer matrix).
Because readers may arrive at this post from anywhere, the prompts should work for someone starting cold — no prior artifacts exist yet in this series, so this is the one week with no catching-up to do, and the post can say so as an invitation.
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. An employed product manager wondering whether restlessness is a signal, a laid-off retail manager with no stay option, and a freelance designer considering a return to full-time 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 — the outline's hook statistics are directional and unverified, so none are being handed to you. If you find yourself reaching for a regret percentage or a layoff figure, that is the signal to restructure the sentence so it does not need one.)
## 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: 1` 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 current hiring trend, layoff figure, salary level, or regret statistic as fact, and none delivers a stay-or-quit verdict. The decision framework organizes the reader's own evidence; market numbers come from the reader or from named sources they check themselves.
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.
Most job searches start with the résumé, which is roughly like starting a road trip by washing the car. This week's three prompts work the ground underneath it: a fast diagnostic that separates a bad month from a bad fit, a translation of what you actually do all day into the language employers hire for, and a full stay-or-go decision memo with your real compensation baseline inside it. All three run on one rule the rest of this series keeps — the AI is your analyst, your researcher, and your sparring partner, never your ghostwriter, because every word a hiring human reads has to be yours and has to be true. AI backstage, you on the page: used that way, there is nothing here a recruiter could ever hold against you.
The Bad Month or Bad Fit Test
Tell a bad month from a bad fit, in twenty minutes.
The worst job moves are not the ones people think about too long. They are the ones made on a Tuesday afternoon after a meeting that went badly, when the fastest available relief is opening a job board. What that impulse skips is diagnosis: the difference between a project that will be over in six weeks and a role that is structurally wrong for the person doing it. Both feel identical at 4 p.m. on the bad Tuesday, and they call for completely different responses. And a word to readers arriving here without a choice — if your last role ended in a restructuring, including one your employer framed around efficiency or AI, none of this is a lecture about whether you should have left. You are here to name what you want to avoid repeating. That is the same diagnostic work, aimed forward.
Why this matters now
The reason this matters right now is timing: the honest audit is nearly impossible to run once you are three applications deep and emotionally committed to leaving. Diagnosis has to come before the search or it becomes a justification of it. This prompt takes twenty minutes and costs nothing, and it produces the one artifact every later week in this series leans on — a clear statement of what is actually wrong. Readers who are already searching can still run it; it just does double duty as a filter for what to say yes to.
You are acting as a career analyst helping me think clearly. You are not a career coach with an agenda, and you will not tell me what to do.
Here is MY SITUATION:
- Current status: employed and restless / recently laid off or ending soon / freelancing or between things / other
- Role and industry:
- How long I have been in this role:
- What set this off — the specific recent thing that made me open this:
- What I have already tried to fix it:
Below is everything that is bothering me, in no particular order. It is unedited on purpose:
[write freely here — a paragraph, a list, whatever comes out]
Do this, in this order.
1. Sort every complaint I listed into one of five buckets: the work itself, the people and management, growth and direction, money and security, energy and health. If something belongs in two buckets, put it in both and say why.
2. For each item, label it Episodic or Structural. Episodic means it plausibly resolves within 90 days without changing employers — a bad quarter, one specific project, a manager who is leaving. Structural means it is built into the role, the company, or the field, and would still be true a year from now. If you cannot tell from what I gave you, label it Unclear and say exactly what you would need to know to decide.
3. Tell me which bucket is carrying the most weight, and say so plainly if one bucket is doing all the work while the others are fine.
4. Ask me the three sharpest questions you still need answered before anyone could reasonably judge whether this is a bad month or a bad fit. Ask them one at a time and wait for my answer before asking the next.
Rules. Do not tell me whether to stay or leave. Do not estimate salaries, hiring conditions, how long a search would take, or how my field is doing — you cannot see that data and I am not asking for it. If I told you I have been laid off, drop the stay-or-go framing entirely and treat this as diagnosing what I need to avoid repeating. Use my own words back to me wherever you can.
Finish with one sentence, in plain language, naming what is actually broken. Then ask me to rewrite that sentence in my own words.
How the AI reads this prompt
Practical examples from different industries
The following are illustrative scenarios, not case studies.
An employed product manager at a mid-size software company.
She pastes in a list built over one bad week: a roadmap reversed twice, a skip-level who does not know what her team ships, no promotion conversation in eighteen months, and a persistent Sunday-evening dread. The sort puts the roadmap churn under the work itself and labels it Episodic — a reorganisation is underway and ends next quarter — while the promotion silence and the invisibility land under growth and direction, both Structural. The output she cares about is the imbalance: her energy complaints are entirely downstream of a growth problem she had been describing as burnout. That reframing changes what she does next, from resting to asking for a scoped conversation about her path.
A retail store manager laid off in a chain consolidation.
He answers the status line honestly, and the prompt drops the stay-or-go framing on the spot. His input is a mix of grief and specifics: eleven years, a store he rebuilt, sixty-hour holiday weeks, and a growing dislike of the district reporting culture. The buckets separate what he lost from what he was already tired of — the people work and the problem-solving sit under the work itself and are things he wants back, while the schedule and the reporting are Structural features of the format, not of that employer. His closing sentence is not about the layoff at all. It names a shift-pattern he will not sign up for again, which is a filter he can use on every posting he opens.
A freelance designer considering a return to full-time.
She has no manager to complain about, so two of the five buckets come back nearly empty, which is itself the finding. Everything she lists clusters in money and security — irregular invoicing, no coverage, a client concentration problem she can name precisely — while the work itself scores clean. The Episodic label lands on a specific slow quarter; the Structural label lands on the concentration. Seeing them separated stops her from converting a cash-flow scare into a permanent decision, and points her at a narrower fix than a full return to employment.
Creative use case ideas
- Deciding whether to leave a volunteer board or committee. The same episodic-versus-structural cut works on unpaid commitments, where guilt usually does the deciding. Swap the five buckets for the work, the people, the mission, the time cost, and the energy.
- A graduate student weighing whether to leave a program. Substitute advisor for manager and funding for compensation. The bad-semester-versus-bad-fit distinction is nearly the same problem with much higher exit costs, which makes the diagnosis more valuable, not less.
- Auditing a long-running creative project. A novel, a band, a podcast three years in. Bucket the complaints and see whether the problem is this chapter or this form.
- A shared version for two people making one decision. Each person runs it separately on the same situation, then compares the five-bucket sorts before talking. The disagreements are the conversation.
- A quarterly self-check-in with no crisis attached. Run it every three months when nothing is wrong, and the trend across four sorts tells you far more than any single one did.
Adaptability tips
Change the five buckets to match what you actually care about — a nurse might split the work itself into patient care and administrative load, which are wildly different experiences filed under one job. Change the 90-day window in the Episodic definition to match your real horizon: a teacher should use the school year, a seasonal retail worker should use the season, and a contractor should use the length of a typical engagement. If you want a shorter run, cut step 4 and keep the sort. If you want a longer one, add a sixth bucket for the commute or the location, which quietly drives more of these decisions than people admit. And if the output feels too gentle, add one line: assume I am underselling the problems in the bucket I mentioned last, and ask about that bucket first.
Pro tips
- Run it twice, a week apart, without rereading the first result, then compare. Items that appear both times are structural regardless of how the model labelled them.
- Paste in an old text or note to a friend from six months ago and add it as evidence. Your past self is a better witness than your present one, and the model will treat it as data.
- If the AI starts giving advice anyway, do not argue with it — reply with one line reminding it of its role, and it will re-anchor. This is faster than restarting.
- Ask for the bucket sort in a strictly neutral register with no reassurance. Some people find the warmth actively obscures the pattern.
Prerequisites
Twenty minutes and a place where you will not be interrupted. Some honesty about the trigger — the specific incident, not the tidy version. Nothing else: no numbers, no documents, no prior artifacts. This is the opening week of the series, so nothing is expected of you that you have not already lived, and that will not be true again after this week.
Required tools
Any general-purpose AI chat tool that holds a back-and-forth conversation. A free tier is sufficient for this variation. The only feature that genuinely matters is that the tool remembers what you said earlier in the same conversation, since step 4 depends on it.
Frequently asked questions
What if the AI tells me to quit anyway?
It sometimes will, because advice is the strongest current in its training data on this subject. Do not accept it and do not debate it — reply with a single line restating that it is an analyst and does not deliver verdicts, and it will drop back into role. If it happens twice in one session, move the rules paragraph to the top of the prompt rather than the bottom; instructions placed early tend to hold better across a long exchange.
I have been laid off. Is this prompt useful, or is it for people with a choice?
It is useful, and the branch in the prompt is there specifically for you. Run this way it stops being a stay-or-go tool and becomes a specification for what you accept next — which parts of the last role you want back, which you will not repeat, and what a bad fit looks like when it is still disguised as an offer. That specification is worth more than it sounds when the third recruiter call starts blurring into the second.
Can it see whether my industry is actually hiring?
No, and the prompt forbids it from pretending otherwise. It has no view of live postings, current conditions in your city, or what happened in your field last month, and anything it produced along those lines would be a confident guess dressed as a fact. Market reality comes from postings you read, official labour statistics you look up, and people who currently do the job. This prompt handles the part that lives entirely inside your own experience.
Everything came back Structural. Does that mean I should leave?
No — it means the situation is unlikely to fix itself, which is a different statement. Structural problems can be addressed by changing the role inside the same company, renegotiating scope, or moving teams, and a whole option gets skipped when Structural is read as an exit signal. Variation 3 exists to weigh those alternatives properly. What you have here is a diagnosis, not a prescription.
Recommended follow-up prompts
- The one-sentence check. Take the sentence you wrote at the end, and ask an AI to interview you for ten minutes about whether it is actually true, with permission to challenge you once per answer. Cheap, uncomfortable, and effective.
- The Market Translation Audit (Variation 2 of this post) — the natural next step once you know what is broken.
- A three-year pattern scan. Feed in three past roles and ask for what the exits had in common, using the same five buckets. If a bucket appears in every exit, you are looking at a pattern rather than a series of unlucky employers.
Tags and categories
Tags:
career audit, job search, decision-making, self-assessment, burnout, stay or go, diagnostic prompts, beginner prompts Categories: Career & Work, Decision-Making Prompts
Citations
NOT APPLICABLE — this variation makes no factual claims requiring an external source. Every output it produces is derived from the reader's own answers.
The Market Translation Audit
Turn what you actually do into what employers actually search for.
There is a specific, maddening gap between what people do all day and what their job title says they do, and job searches are lost inside it. The operations manager who is really running vendor negotiation, the teacher who has been doing instructional design for four years, the freelancer who has quietly been a client strategist — all of them get filtered out by systems and humans scanning for words they never learned to use about themselves. The fix is not inflation and it is certainly not a thesaurus pass over the résumé. It is translation: taking the concrete thing you did on Tuesday and finding the phrase the market uses for it, then holding onto the evidence that proves you did it. That evidence is the part everyone drops, and it is the part that survives an interview.
Why this matters now
This is the week to build the inventory because everything after it depends on the vocabulary. Week 2 defines a target you cannot define without knowing what you are selling; the résumé and outreach weeks are pure editing work on top of this list; the interview and negotiation weeks are you defending these specific claims out loud. Building it now, before urgency compresses your judgment, means you are choosing your framing rather than reaching for whatever sounds impressive at 11 p.m. And doing it while you can still remember last week is genuinely easier than reconstructing it from an old job description in four months.
You are a market translator. Your job is to convert what I actually do into the language employers use when they write job descriptions and recruiters use when they search — without inflating anything.
MY SITUATION:
- Current or most recent role and industry:
- Roughly how long I have done this kind of work:
- Where I think I might go next, if I have a guess (write unsure if you do not):
- The audience I want to be legible to (for example: hiring managers at mid-size software companies, retail operations directors, design studios):
Here is my raw material. It is deliberately unpolished — I am giving you the actual work, not the résumé:
- Things I did in the last two weeks:
- Things people come to me for that are not in my job description:
- Things I fixed that stayed fixed:
- Tools, systems, and software I touch weekly:
- The parts of my work I would do for free:
- The parts that drain me:
Produce four grouped lists, in this order and no other.
GROUP 1 — CAPABILITIES. For each one, give three things on one line: the plain-English thing I do, then the phrase my target audience would use for it, then the specific evidence from my own material that supports it. Where the evidence is weak, write Thin evidence and tell me what proof I would need to make that claim defensible in an interview.
GROUP 2 — DURABILITY READ. Sort my capabilities into Likely appreciating, Likely steady, and Likely commoditising. Give one sentence of reasoning for every placement, and mark each one clearly as your reasoning from general patterns rather than from data. You cannot see the current job market. Do not state hiring demand, salary levels, or industry trends as fact. Where a placement depends on something checkable, name what I should verify and where I should look — postings in my target market, official labour statistics, or someone currently doing the job.
GROUP 3 — GAPS. What is missing from my material that my target audience will expect to see, and the cheapest honest way to close each gap within 90 days.
GROUP 4 — QUESTIONS FOR ME. The five questions whose answers would most change your read.
Constraints. Use only my material — do not invent achievements, metrics, team sizes, or scale I did not give you. If a number would strengthen a line, ask me for it rather than supplying a plausible one. Do not write anything in first person as though it were a finished résumé line; give me the raw phrasing and I will write the line myself. Keep every market phrase to something a human would say out loud without wincing.
How the AI reads this prompt
Practical examples from different industries
Illustrative scenarios.
A middle-school science teacher exploring corporate learning roles.
Her raw material is entirely classroom: unit planning, differentiating for a class with eight reading levels, running a department-wide rollout of new lab equipment, and being the person everyone asks about the grading system nobody understands. The translation surfaces curriculum design, needs assessment, stakeholder training, and change rollout, each tied back to a specific term or project rather than asserted. The durability read flags her assessment-design work as likely appreciating and marks it as reasoning, not data, pointing her at postings in corporate learning teams to verify. The gap list is short and specific: no vocabulary for measuring outcomes in business terms, closable by reframing two projects she already ran.
The laid-off retail store manager, now building an inventory.
He has never written any of this down, and his first instinct is that eleven years of retail does not translate. The prompt's raw-material questions do the work — shrink reduction he can attribute to a process he designed, a scheduling system he built that outlasted him, the fact that three district peers called him about staffing problems that were not his. What comes back is inventory control, workforce scheduling, loss prevention process design, and a demonstrable pattern of informal peer authority, each carrying an evidence line. Two claims come back marked Thin evidence, which tells him precisely which numbers to request from a former colleague before he uses them.
A freelance designer preparing for full-time conversations.
Her material is fragmented across a dozen clients, which is the classic freelance problem: enormous range, no legible arc. The grouping produces something she had not seen — that the same three capabilities recur across every engagement regardless of the deliverable, and that the varied client list is evidence of adaptability rather than a lack of focus. The gaps group is the useful part: her material contains almost no language about working inside a team or a design system, which is exactly what her target audience screens for. She spends the following month building that evidence rather than redesigning her portfolio again.
Creative use case ideas
- A parent returning to work after years of caregiving. Run it on the caregiving years directly: scheduling under constraint, budget management, coordinating across institutions, advocacy in adversarial systems. The evidence requirement keeps it honest, which is what makes it usable rather than defensive.
- Military-to-civilian translation. The structure is built for exactly this problem — a highly specific internal vocabulary that outsiders cannot parse — and the evidence line prevents the flattening that makes so many translated résumés read as vague.
- Writing a grant application bio for a community organisation. Same mechanics, different target audience. Swap hiring managers for a specific funder and the phrasing shifts accordingly.
- Preparing to hand off a role you are leaving. Run it on your own job and the capabilities list becomes a handover document and a job description for your replacement, which is a genuinely kind thing to leave behind.
- A hobby you are considering going semi-professional with. The gaps group is brutal and useful here, because hobbies accumulate skill unevenly and you rarely notice which part is missing until a paying client asks for it.
Adaptability tips
The two-week window in the raw material block is a starting point, not a rule — use the last quarter if your work is project-shaped and the last two weeks would be unrepresentative. The four groups can be run separately across four sessions if the output is overwhelming; Group 1 alone is a complete deliverable. To sharpen the durability read, add your target audience's actual job postings as pasted input and ask the model to map your capabilities against the language in them, which converts a general pattern into a specific comparison against real listings you found yourself. To go narrower, name a single target role rather than an audience. To go broader, run it twice with two different target audiences and compare which capabilities survive both translations — those are your portable ones.
Pro tips
- Feed in your current job description as a separate item and ask what you do that it does not mention. The delta is usually where your next role lives.
- Ask for the market phrasing in two registers: the formal one for written applications, and the spoken one for conversations. They differ more than you would expect, and using the written register out loud sounds rehearsed.
- After Group 4, actually answer the five questions and rerun. The second read is materially better and most people never do it.
- Ask the model to flag any phrase it produced that it thinks is currently overused. It is unreliable about what is current, but it is useful at spotting the phrasing that reads as filler.
Prerequisites
Thirty to forty minutes and access to something that jogs your memory — a calendar, a task board, a project list, an inbox. Ideally the diagnosis from Variation 1, since knowing what is broken helps you notice which capabilities you actually want to keep using. If you have a target audience in mind, bring it; if you do not, write unsure and the prompt still works, it just returns a broader read.
Required tools
Any general-purpose AI chat tool. A free tier handles this comfortably. If your tool supports uploading a document, uploading your current job description or a project list alongside the prompt improves the input quality noticeably — but pasting the same material as text works identically well.
Frequently asked questions
Is this not just résumé keyword stuffing with extra steps?
No, and the difference is the evidence line. Keyword stuffing attaches impressive words to a person and hopes nobody checks; this attaches market vocabulary to a specific thing you did and keeps the proof bolted to the claim. The output is deliberately not résumé-ready, because a claim you cannot support in a twenty-minute conversation is a liability rather than an asset. If anything, this process removes more claims than it adds.
The durability read said one of my main skills is commoditising. Should I panic?
No — read the marking on it. The prompt requires the model to label that judgment as reasoning from general patterns rather than data, because it cannot see what is happening in your field right now. Treat it as a hypothesis worth testing against real postings, official labour statistics, and people currently doing the work. A prediction from a model with no view of the present is a prompt to go look, not a finding.
What if my raw material is genuinely thin?
Then the output will tell you so, in the form of Thin evidence markers and a long Gaps list, and that is a real result rather than a failure. Thin material usually means the memory is the problem rather than the work — most people cannot reconstruct a quarter from memory. Go back with a calendar or a task history and rerun; the second pass typically doubles the input.
Can I use the phrases it gives me directly in my résumé?
Use them as vocabulary, not as sentences. The prompt specifically asks for raw phrasing rather than finished lines because material that arrives pre-written tends to stay pre-written, and hiring humans have become good at spotting the seams. Take the phrase, put it into your own sentence, and make sure you can say the sentence out loud without hesitating. That last test catches almost everything.
Recommended follow-up prompts
- The posting gap read. Paste in five real job postings you found yourself and ask the model to map your capabilities list against the language they actually use, flagging what appears in all five and is missing from your inventory.
- The Stay-or-Go Decision Memo (Variation 3 of this post) — the inventory feeds directly into the options section.
- The reverse interview. Ask an AI to act as a skeptical hiring manager and challenge each capability in your list once, demanding the evidence. Uncomfortable, and much cheaper than discovering the weak claim in a real interview.
Tags and categories
Tags:
skills inventory, career translation, transferable skills, résumé preparation, job search, market positioning, intermediate prompts, evidence-based Categories: Career & Work, Prompt Design
Citations
- U.S. Bureau of Labor Statistics, Occupational Outlook Handbook — bls.gov/ooh. Referenced as the official source readers should check for occupational information rather than asking an AI to assert it.
- O*NET OnLine, sponsored by the U.S. Department of Labor, Employment and Training Administration — onetonline.org. A public database of occupational skill and task descriptions, useful for checking the standard vocabulary attached to a given occupation.
The Stay-or-Go Decision Memo
A decision memo your future self can hold you to.
Six months from now you will not remember why you decided this. You will remember the feeling, which will have quietly rewritten itself to match whatever happened next, and you will have no way to tell a good decision that got an unlucky outcome from a bad decision that got away with it. That is the case for writing it down. A decision memo is a boring artifact that does three unglamorous things: it fixes your reasoning in place before the outcome contaminates it, it forces the money out of the vague into the specific, and it gives you a date to look at it again rather than relitigating it every Sunday night. This is the budget week of the series, and like every budget week, the discomfort is the point — most people cannot say what they currently earn once benefits, vesting, and the things that vanish on resignation are counted.
Why this matters now
Do this before you talk to anyone, because a compensation baseline built after a recruiter has anchored you is not a baseline, it is a reaction. This is also the week that pays off latest: the number you produce here is what Week 7's negotiation stands on, and the criteria you weight here are what Week 8's offer comparison reuses. There is a nearer-term payoff too. The runway figure — how long you could go without income at your actual floor spending — changes how a search feels from week one, because urgency you have measured behaves very differently from urgency you have only felt.
We are going to produce a decision memo that I write to myself about whether to stay, grow in place, or go. You are my analyst and my interviewer. You are not my advisor, and you will not deliver a verdict — the decision line at the end is mine to write, and this memo is worthless if it turns out to be yours.
Run this in four phases. Do not skip ahead, and do not produce the memo until Phase 3.
PHASE 0 — INTAKE. Ask me for MY SITUATION, then wait for my answer before continuing:
- Status: employed and considering a move / laid off or ending soon / self-employed or freelance considering a change
- Role, industry, tenure, and any location constraints
- Who else this decision lands on
- The date by which this has to be decided, and what is forcing that date
PHASE 1 — STRUCTURED INTERVIEW. Interview me in five rounds. One round per message, maximum four questions per round, and wait for my answers before moving to the next round.
Round A — the trigger and the history: what changed, and how many times I have felt this before.
Round B — the money, fully loaded: base pay, variable pay and when it actually pays out, employer retirement contributions and their vesting schedule, health coverage and what it costs me, equity and its dates, anything that disappears the day I resign, and anything that only pays if I stay past a specific date.
Round C — runway: liquid savings, my monthly floor spending, other household income, dependants, and how long I could go without income before this decision starts making itself.
Round D — the non-financial ledger: what I would miss, what I would escape, what I am afraid of, and what I would regret in five years in each direction.
Round E — the alternatives I have not taken seriously: an internal move, a scoped conversation with my manager, a change of team, a sabbatical, or a smaller change that is not a job change.
If an answer is vague, push once. If I say I do not know a number, mark it UNKNOWN and carry it forward as an open gap — do not estimate it for me.
PHASE 2 — CRITERIA AND WEIGHTS. Propose six to eight decision criteria drawn from what I actually said rather than from a generic list. Show them to me, let me cut, add, and rename them, then ask me to assign weights totalling 100. Do not assign the weights yourself.
PHASE 3 — THE MEMO. Write it in these sections, in this order.
1. Situation — three sentences, my facts only.
2. Evidence — what I told you, grouped, with every UNKNOWN listed plainly as an open gap.
3. Financial baseline — my current compensation decoded into one fully loaded figure with its components shown, my runway in months at my stated floor, and the specific dates on which money or coverage changes if I leave. Use only numbers I gave you. Where a number is missing, write UNKNOWN and state what it would change.
4. Options — Stay, Grow in place, Go, plus any fourth option my answers surfaced. Score each against my weighted criteria, show the arithmetic, and state what each score is most sensitive to.
5. Red team — argue the strongest honest case against whichever option scored highest, and then against whichever option I seem emotionally attached to. Name that attachment if you saw it.
6. What would change my mind — the specific, checkable events that should flip this decision.
7. Decision — leave this section blank, with a prompt for me to fill in, in my own words.
8. Revisit date — propose one, and say what should be true by then.
PHASE 4 — HANDOFF. List the three artifacts this memo produces: my written decision, my compensation baseline, and my search timeline. In one line each, tell me what I will need them for later.
Standing rules for every phase. Do not state current hiring conditions, layoff patterns, salary benchmarks, or how long a search takes as fact — you cannot see that data. Where the decision turns on market reality, name the source I should check and exactly what to look for in it. This is not financial, legal, or tax advice and you should not frame it as such; where a benefits, equity, or severance question needs a real answer, tell me to get it from my plan documents, my HR portal, or a qualified professional. Everything you produce here is a draft I will rewrite in my own words, and none of it is written for anyone to read but me.
How the AI reads this prompt
Practical examples from different industries
Illustrative scenarios.
The product manager with unvested equity and a spring bonus.
Round B pulls out what she had been carrying vaguely: a bonus that pays in March, a retirement match with a cliff eleven months out, and an equity tranche vesting in between. The financial baseline decodes her fully loaded compensation into a figure meaningfully higher than the salary she had been mentally comparing against offers, and it produces something she had never seen — a dated calendar of what leaving costs in each of the next four months. She still leaves. But the memo shifts her timeline by seven weeks, and the red team section names her attachment to the go option out loud, which she reads twice.
A reader whose role ends in six weeks with severance attached.
The stay option does not exist, so the memo restructures around a different question: accept the first workable offer, or hold out for the targeted one. Round C does the heavy lifting, because severance plus savings against actual floor spending produces a runway in months, and that number sets how long holding out is affordable rather than how long it feels affordable. The memo marks two entries UNKNOWN — the exact end date of his health coverage and whether his severance affects unemployment eligibility — and routes both to his HR portal and his state agency rather than answering them. The revisit date lands at the runway midpoint, with a named condition attached.
A dual-career household weighing a relocation.
The intake question about who else this decision lands on stops being a formality, and the criteria that emerge in Phase 2 include two that no generic list would contain: a partner's licensure that does not transfer across state lines, and a school-year boundary. Both get weighted heavily, and both change the option scores substantially. The most useful output is the sensitivity note — the memo shows that the entire comparison turns on one unknown about the partner's credential, which converts a sprawling months-long argument into a single question worth answering this week.
Creative use case ideas
- Deciding whether to keep or sell a rental property, a small business, or an inherited asset. Same structure, different rounds: the money round becomes carrying costs and tax dates, and the emotional-attachment red team is, if anything, more necessary.
- A nonprofit board deciding whether to sunset a program. Run Phase 2 collectively so the criteria and weights are negotiated in the open. Most board deadlocks are undeclared weight disagreements wearing a factual costume.
- Choosing between graduate programs, or whether to go at all. The runway round translates directly, the vesting round becomes funding and stipend dates, and the revisit date is genuinely useful when the decision is reversible in year one.
- A creative deciding whether to keep going on a long project. Replace compensation with time, energy, and opportunity cost. The red team section is the part that earns its keep.
- A household deciding whether to move cities for reasons that have nothing to do with work. The weighted criteria step gives two people a structured way to discover they have been optimising for different things.
Adaptability tips
If five rounds is more than you have patience for, cut Round E and run it separately later — it is the round most people skip and the one that most often produces a fourth option. If you are running this with a partner, run Phase 1 separately and merge at Phase 2, where the weight negotiation happens; merging earlier means the louder person's framing sets the questions. For a lighter version, keep Phases 0, 2, and 3 and supply the interview material as a written brief instead of answering rounds. To make it reusable, save the criteria and weights you settled on and reuse them in Week 8 against real offers — that continuity is most of the value. And if your tool loses the thread partway through, paste a summary of your answers so far back in and name the phase you were in; long protocols survive interruption better when you carry the state forward explicitly.
Pro tips
- Do Round B with the actual documents open. Almost nobody has their vesting schedule memorised correctly, and a memo built on a misremembered cliff date is worse than no memo.
- Ask for the red team section twice, the second time with the instruction to be less polite. The gap between the two versions is informative.
- Before Phase 2, write down your own six criteria on paper without showing the model. Compare. What you left off is usually what you have been avoiding.
- After the memo, ask for a one-paragraph version you could read in thirty seconds. Save both. The long one is for now; the short one is what you will actually reread at the revisit date.
Prerequisites
Ninety minutes, ideally in two sittings, and your actual numbers — recent pay statements, your benefits summary, your equity or retirement plan documents, and a realistic figure for monthly floor spending. Variations 1 and 2 of this post feed it directly: the diagnosis supplies the trigger and the inventory supplies the options. Do not run this from memory. The whole value is in the specificity, and estimated inputs produce a memo that reads authoritative and is not.
Required tools
Any general-purpose AI chat tool that maintains a long conversation without losing earlier context — this protocol runs across many exchanges and depends on the model remembering Round B when it writes section 3. If your tool struggles with long sessions, run it in two conversations and paste a summary of Phase 1 answers into the second. A document you can paste the finished memo into and keep. No paid tier is strictly required, though longer context makes it smoother.
Frequently asked questions
This is a lot of work for a decision I might make anyway. Is it worth it?
The work is the point, and most of it is work you would otherwise do badly and repeatedly. People re-decide this question dozens of times over months without ever writing it down, which costs far more hours than one structured session and produces nothing durable. The memo also does something rumination cannot: it fixes your reasoning before the outcome rewrites it, which is the only way to learn anything from this decision for the next one.
What if the scoring comes out somewhere I do not like?
Then you have learned something about your weights, not about your future. A weighted score is a mirror of the values you assigned in Phase 2, so a result that feels wrong usually means a weight is wrong — and noticing that is genuinely useful information. Go back, adjust the weight you now realise was understated, and rerun. The decision line is still blank and still yours; the arithmetic has no authority over you.
Should I trust the runway number?
Trust it exactly as far as you trust the floor-spending figure you supplied, which for most people is optimistic on the first pass. Pull three months of actual statements rather than estimating, and include the annual and irregular costs that do not show up in a typical month. The prompt marks anything you did not know as UNKNOWN specifically so those holes stay visible rather than getting absorbed into a total.
Can the AI tell me whether my severance or benefits situation is normal?
No, and it is instructed to route you elsewhere when that comes up. Severance terms, benefit continuation, equity treatment on departure, and unemployment eligibility are governed by your specific plan documents, your agreement, and your jurisdiction — none of which the model can see, and all of which it can produce confident-sounding wrong answers about. Get those from your plan documents, your HR portal, or a professional. The memo's job is to hold the answer once you have it.
Do I have to share this with anyone?
No, and the prompt explicitly states nothing in it is written for another reader. That is deliberate. A memo written with an audience in mind stops being honest around section 5, which is exactly where the honesty was supposed to be.
Recommended follow-up prompts
- The revisit prompt. On your revisit date, paste the memo back in and ask what has actually changed, what you predicted correctly, and whether the conditions you named have been met. This is where the artifact starts paying compound interest.
- The timeline builder. Convert the memo's decision and dates into a week-by-week search timeline with named checkpoints — the third artifact this week produces, and the one Week 2 picks up.
- The offer matrix (Week 8 of this series) — the same criteria and weights, applied to real offers instead of hypothetical options.
Tags and categories
Tags:
decision memo, stay or go, compensation baseline, runway, career decision framework, weighted criteria, red teaming, advanced prompts, job search Categories: Career & Work, Decision-Making Prompts, Prompt Design
Citations
- U.S. Department of Labor, Employee Benefits Security Administration — An Employee's Guide to Health Benefits Under COBRA, dol.gov. Referenced as the authoritative starting point for readers assessing what health coverage continuation would cost them after leaving a job.
- HealthCare.gov — Special Enrollment Period guidance for loss of health coverage, healthcare.gov. Relevant to the coverage-date questions the memo's financial baseline section surfaces.
- U.S. Department of Labor, Employee Benefits Security Administration — What You Should Know About Your Retirement Plan, dol.gov. Referenced for readers checking vesting schedules and employer contribution terms in their own plan documents rather than asking an AI to characterise them.
Which of the three should you use?
The three prompts answer three different questions, and running them in order is not an accident. Variation 1 asks what is actually wrong, and it is the only one you can run with no preparation, no numbers, and no documents — twenty minutes and some honesty about the trigger. Variation 2 asks what you are working with, converting a work history into the vocabulary the market hires for and the evidence that survives an interview. Variation 3 asks what you should do about it, and it is the only one that produces a decision, a number, and a date. Skipping to Variation 3 is possible but usually worse, because its options section is only as good as the diagnosis and the inventory underneath it.
They also teach three different prompting skills, which is why they are not one prompt at three lengths. Variation 1 is about taxonomy and the discipline of a forced binary — bringing your own categories rather than letting the model invent flattering ones, and giving it a clean way to say it does not know. Variation 2 is about shaping output: named audiences, explicit group structure, and constraints that block the model's inflation reflex. Variation 3 is about protocol design — phases, turn-taking, evaluation criteria you define, weights you assign, and an adversarial review step aimed at the conclusion you are most inclined to like. If you take one transferable idea from this post, take that last one: the moment you decide who assigns the weights, you have decided who is making the decision.
Choose by where you actually are. If you are in the bad-Tuesday state and about to open a job board, run Variation 1 and nothing else today. If you have already decided to look and want to move, run Variation 2 this week and Variation 3 before you talk to anyone. If you were laid off, run Variation 1 with its branch, then Variation 3 restructured around accept-now versus hold-out — the stay column will be empty and the rest of the memo works exactly as well. And whichever you run, the parts that go out into the world afterward get written by you. The AI stays backstage.
TAGS:
