Eight Weeks, One Reusable System: Landing the Plane
WEEK 99 :: 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: "Landing the Plane" — Post-Trip Reconciliation and the Reusable System.
This is Week 8, the final week of an eight-week series on planning a vacation with AI. The trip is over. This week is about closing it out well and, more importantly, about keeping what the reader just built — turning eight weeks of one-off prompts into a personal, reusable system so the next trip takes a fraction of the effort. This is the payoff the whole series was pointing at.
There are two jobs. The first is reconciliation: a budget post-mortem comparing what was planned against what was actually spent and finding where the estimates were wrong; and the recovery work of disputing incorrect charges and pursuing any compensation the reader is owed — a delayed flight, a resort fee that was never disclosed, a charge that does not match what was agreed. The second job is the durable one: taking the workflow the reader has now lived through and compressing it into a template they can run again, so the knowledge does not evaporate the moment they unpack.
The deliverable the reader should walk away holding is twofold: a clear post-trip reconciliation — what was spent versus planned, what to dispute and how — and a reusable personal trip-planning template distilled from the eight-week process, ready to run for the next destination.
The three prompts should help a reader:
- Run the budget post-mortem. Compare planned against actual from the reader's own records, surface where the plan was optimistic, and turn that into a sharper set of assumptions for next time — the AI structuring the comparison from numbers the reader supplies, not inventing what a trip "should" cost.
- Pursue disputes and claims, methodically. Organise a charge dispute or a delay claim into who to contact, what evidence to attach, and what to ask for, in a calm and orderly sequence — while pointing the reader to the airline, card issuer, or platform to confirm what they are actually entitled to rather than asserting it.
- Build the reusable template. Distil the eight-week workflow into a personal, repeatable planning system — the steps, the prompts worth keeping, and the reader's own hard-won preferences — so the next trip starts from a framework instead of a blank page. This is the series' real lesson: a good prompt, saved and adapted, becomes a system.
At the advanced tier, the strongest version of this week is a written, reusable planning template the reader can save and re-run — the whole series compressed into a sequence of steps and prompts tuned to how this traveller actually plans, plus a reconciliation summary that feeds next time's estimates. That structure is worth reaching for, and it is the natural landing point for a series about turning one-off AI help into a repeatable method.
A hard constraint, carried through to the last week. AI models cannot see the reader's actual receipts, a specific carrier's current compensation policy, or the rules that govern a particular claim, and these are jurisdiction- and date-specific. No prompt may ask the AI to confirm a specific compensation entitlement, promise a payout, adjudicate whether a charge is disputable, or state current claim rules as settled fact. The dispute and claim prompts should give the reader an organised process and the right questions — and send them to the airline, card issuer, or regulator to confirm — not a verdict on their case.
Design the prompts so the AI does what it is genuinely good at: structuring a reconciliation from supplied numbers, organising a dispute into an orderly sequence, and compressing a lived process into a reusable template. The reader supplies their records and their experience; the AI supplies structure, sequence, and the distilled system. Posts whose prompts have the AI assert entitlements or invent costs should expect to be marked down on Practical Utility and Content Accuracy.
Series dependency chain, for the Metadata block: Week 8 consumes the entire trip — the budget ceiling from Week 1 (to reconcile against), the bookings and protection work from Weeks 3, 4 and 6 (for disputes and claims), and the in-trip record from Week 7 (what actually happened and what it cost). Week 8 produces the reusable planning template — the series' closing artifact — which has no successor week because it is the thing designed to start the next trip's Week 1.
Because readers may arrive at this post having only just returned, without having read the earlier weeks, the prompts should work for someone who simply has their receipts and a sense of how the trip went, while making clear the reusable template is far richer when it is distilled from the full process.
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 consumer travel topic. The template lists tech startup / retail / freelance as suggested industry examples — those are marked MAY, and this week you should almost certainly adapt them. A family reconciling a trip's spending against the plan, a traveller pursuing a delayed- flight claim, a couple disputing an undisclosed resort fee, and anyone building a personal template so the next trip is easier are the right contexts here. 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. Given the claims-and-entitlements constraint above, this is a bad week to invent any — if you find yourself reaching for a compensation amount or a typical trip cost, that is the signal to restructure the prompt so the reader supplies the real number and confirms entitlements at the source.)
## 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: 8` 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 confirm a compensation entitlement, promise a payout, adjudicate whether a charge is disputable, or state current claim rules as settled fact. Those must be a process and questions the reader confirms with the airline, card issuer, or regulator.
You are home, the laundry is running, and there is a small pile of evidence on the kitchen counter: receipts, a boarding pass from a flight that left four hours late, and a card charge nobody in the house recognises. This week closes the series with three prompts — a plain budget post-mortem that shows exactly where your plan was wrong, an orderly way to assemble a dispute or delay claim without spiralling, and a builder that compresses these eight weeks into a personal planning playbook you can run again. The first two recover money and clarity. The third is the one that makes the next trip cost a fraction of the effort, which is what the whole series was pointing at.
The Honest Ledger
Find where your trip budget lied, using only your own receipts.
Every trip has a moment, usually somewhere around the second week home, when the card statement arrives and the number is not the number you had in your head. It is rarely one catastrophic line. It is thirty small ones: the airport lunch that cost what dinner should have, the taxi you took because the bus was confusing, the entry fee you had genuinely forgotten existed. The instinct is to look away, because the trip was good and the money is spent and what is the point. The point is that the same gap will open next time, in the same places, for the same reasons — unless you look at it once, calmly, while the memory is still fresh enough to explain the numbers. This prompt is a fifteen-minute reckoning that turns a vague sense of "we went over" into three sentences you can reuse.
Why this matters now
Right now, days after getting back, you have something you will never have again: you can still remember why each charge happened. In six weeks the statement will be a wall of merchant names and the story behind each one will be gone. Doing the comparison now — while you can still say "that one was the day it rained and we bailed on the hike" — is the difference between data and a lesson. It also matters practically: a post-mortem run this week is what makes an incorrect charge visible in time to dispute it, which is exactly what the next prompt picks up.
You are helping me run a simple post-trip budget review. I will paste my planned budget and what I actually spent.
Work only from the numbers I give you. Do not estimate, fill in, or guess at any figure I have not supplied, and do not tell me what a trip like mine 'usually' costs. If something looks missing or unclear, ask me for it instead of inventing it.
Here is what I planned to spend, by category:
[paste your planned amounts]
Here is what I actually spent, by category:
[paste your actual amounts]
Please do three things, in this order:
1. Go category by category. For each one, show what I planned, what I spent, the difference, and whether I was over or under. Then give me the overall total the same way.
2. Name the three categories where I was most wrong in dollar terms, largest first. For each, ask me one short question that would help you tell whether the miss was a one-off event or something I am likely to repeat.
3. After I answer, write three specific assumptions I should carry into my next trip. Write them as sentences I could paste directly into a future budget, built from my actual numbers — for example, 'assume [my real daily figure] per day for food when I am somewhere I do not cook,' not a general recommendation.
Keep it plain and unsentimental. Do not tell me how I should have travelled.
How the AI reads this prompt
Practical examples from different industries
A family of four, one week at the beach.
They budgeted $3,200 and the card says $4,050. Pasting both sets of numbers in, the biggest gap is not the rental house — that was booked months ago at a fixed price and came in exactly as planned. It is "food and everything else," which ran nearly double. The follow-up question the model asks is whether they cooked as often as they expected. They did not: the kitchen was poorly equipped, so a plan built on five dinners in became two. The assumption they carry forward is not "spend less on food." It is "verify the kitchen before budgeting meals at home prices," which is a check they can actually perform next time.
A couple, ten days across two European cities.
Their planned and actual totals are almost identical, which initially reads as a success and is why the category-by-category step matters. Underneath, transport ran 60% over and activities came in far under. The story is that they walked less and taxied more than they expected because one of them had a bad knee by day three, and then had less energy left for the museums they had paid to prioritise. The reusable assumption is about pace rather than price: budget for the transport a realistic body needs, and buy fewer timed tickets in advance.
A solo traveller on a two-week road trip.
Fuel, tolls, and parking were tracked loosely as one line and came in 40% over. The model's question — one-off or repeatable — surfaces that most of the overage was parking in three cities, not fuel across the whole route. The resulting assumption is granular and genuinely useful: split driving costs into "moving" and "parked," because the moving number is predictable from distance and the parked number is a function of which cities are on the itinerary. That is a distinction no generic budget template would have offered.
Creative use case ideas
- The group trip settlement. Six friends, one shared house, a chaotic spreadsheet, and the awkward conversation nobody wants to start. Run the comparison on the group's pooled plan versus pooled actuals first, so the discussion is about categories rather than about who ordered the second bottle.
- The wedding post-mortem. Not for the couple — for whoever is planning next. The same planned-versus-actual pass over a wedding budget produces assumptions that are gold to the next person in the family who gets engaged, and it is the sort of knowledge that otherwise dies with the spreadsheet.
- The semester abroad debrief. A student comparing the cost estimate the programme provided against what four months actually cost, category by category. The output is a genuinely useful document for the next cohort, and it is far more honest than the official figure.
- The holiday season audit. Gifts, travel, food, hosting. Run it in January while the receipts are still findable, and the three assumptions you generate are the only realistic defence against repeating it in December.
- The conference-travel review for a small team. Two people went, the budget was approved on a guess, and the finance question next quarter will be whether to send two people again. A clean planned-versus-actual makes that a five-minute answer instead of an argument.
Adaptability tips
The easiest scaling lever is granularity. If your categories are broad — "food," "transport," "stuff" — the output will be broad, and the assumptions will be too general to act on. Split the categories that hurt and merge the ones that did not: after the first run, most people find one category that deserves three sub-lines and three that could be a single row. That refinement is itself the point of running it.
If you did not keep a plan at all, the prompt still works in reverse. Paste only the actuals and ask the model to organise them into categories and tell you what a plan built on these numbers would look like. You lose the variance analysis and keep the baseline, which is worth having for a first-time traveller with no history to compare against.
For multi-currency trips, state the conversion approach in the prompt rather than letting the model choose: give it the rate you actually paid, or paste your home-currency statement amounts and say so. Otherwise the model may apply a current rate to a purchase made three weeks ago, and you will be reconciling against a number that was never charged to you.
For a household that travels several times a year, keep the outputs. Three trips of assumptions stacked next to each other stop being anecdotes and start being a personal cost model — and that model is exactly the raw material the advanced prompt below is designed to consume.
Pro tips
Paste your actual card statement text rather than a hand-typed summary where you can. Retyping is where the memory-flavoured errors creep in, and the raw merchant lines often jog your recollection of what a charge actually was.
Run the whole thing before you file the receipts, not after. A category that comes in wildly over is frequently a duplicate charge or a fee you did not agree to, and the review is the cheapest fraud check you will ever run.
When the model asks its one-off-or-repeatable question, answer it honestly rather than defensively. The value of the exercise is concentrated entirely in those two or three answers, and a flattering answer produces a useless assumption.
If you travel with someone, do the comparison together and answer the questions out loud. The disagreements about why a category blew up are usually more informative than the numbers.
Prerequisites
You need your planned budget in whatever form it exists — a spreadsheet, a note on your phone, or a number you agreed on out loud and can reconstruct. You also need your actual spending, which for most people means a card statement covering the trip dates plus a rough figure for cash. Nothing needs to be tidy or complete; the prompt is built to work with approximations as long as you say which figures are approximate. Fifteen uninterrupted minutes is genuinely enough.
Required tools
Any general-purpose AI chat tool on its free tier is sufficient — this prompt involves no file handling, no browsing, and no long context. If your statement is a PDF and your tool accepts file uploads, that saves retyping, but copying the text into the chat works just as well.
Frequently asked questions
What if my planned budget was never written down?
Reconstruct it roughly and say so in the prompt — "these planned figures are from memory and are approximate" is a perfectly good input. The model will treat them as soft numbers if you tell it they are soft. A remembered plan compared against a documented actual still surfaces the big misses, which is where nearly all of the value sits.
Will the AI get my arithmetic right?
Mostly, and you should still spot-check the totals. Language models have historically been unreliable at long chains of arithmetic, and while modern tools are considerably better, a review whose numbers you have not verified is not a review. Check the overall total and the largest single variance; if those two are right, the rest is very likely fine.
Should I include flights and accommodation if I booked them months ago?
Include them, and consider marking them as "pre-paid" in your input. They usually come in exactly on plan, which is useful information — it tells you that your fixed costs are predictable and that your variance lives entirely in daily spending. That is a genuinely different problem from booking badly, and it changes what you do next time.
What if the review shows I came in under budget?
Then the interesting question is whether you underspent or overplanned, and the model's follow-up question should tease that apart. Coming in under because you skipped things you had wanted to do is a planning failure wearing a success costume. Coming in under because your estimates were conservative is worth knowing too — it means you can afford to be more ambitious next time.
Recommended follow-up prompts
Run the Week 1 feasibility-and-budget prompt again for your next destination, feeding it the three assumptions this review produced. That is the cleanest demonstration of what this whole series is for: the output of the last trip becomes the input of the next one.
Try a "hidden cost audit" prompt — paste the same actuals and ask the model to flag every line that looks like a fee, surcharge, or add-on rather than a purchase. It is a fast way to find the resort fee you are about to dispute in Variation 2.
If you travel with others, follow up with a cost-splitting prompt that takes your reconciled categories and produces a fair settlement, including the shared costs that one person happened to put on their card.
Tags and categories
Tags:
post-trip review, budget reconciliation, travel budgeting, planned vs actual, personal finance, beginner prompts, variance analysis, trip debrief
Categories:
Travel Planning, Personal Finance
Citations
NOT APPLICABLE
The Calm Claim
Turn a bad travel experience into an organised, sendable case.
There is a particular kind of exhaustion that comes with being owed something. The flight was late, the fee was never disclosed, the charge does not match what was agreed — and you know, roughly, that you have a case. What you do not have is the energy to work out who to email first, what to attach, how to phrase it without sounding unhinged, and what happens if nobody replies. So the thing sits in a drawer for six weeks and then it is too late, which is a quiet outcome that a great many companies are entirely comfortable with. The bottleneck is almost never the merits. It is the organisation. This prompt does the organising, without pretending to know things it cannot know about your particular case.
Why this matters now
Claims and disputes are governed by clocks. Card issuers, airlines, hotels, and booking platforms all work to windows measured in days or weeks from the transaction or the incident, and those windows vary by country, by card network, by carrier, and by what exactly went wrong. The week you get home is the week those clocks are all still running. Getting the sequence right now — who you contact first, what you attach, what you ask for — is worth more than a stronger argument sent three months late. It is also the moment your evidence still exists: photos on your phone, emails in your inbox, a boarding pass in a wallet app that you have not yet cleared out.
Act as a methodical claims organiser helping me prepare one travel dispute. You organise process and evidence. You do not decide outcomes.
The situation: [describe in a few sentences what happened — the charge, delay, or fee, with dates and amounts]
Who I dealt with: [airline, hotel, booking platform, tour operator, or other]
How I paid: [credit card, debit card, booking platform, cash]
Where this happened and where I live: [country or countries]
What I already have: [list your evidence — booking confirmation, receipts, emails, screenshots, photos, boarding passes]
What I want: [a refund of a specific amount, reimbursement of costs incurred, removal of a fee, or a written explanation]
Rules you must follow:
- Do not tell me whether I am entitled to compensation, how much I am owed, or whether this charge is disputable. You do not have my ticket conditions, my card agreement, or the rules that currently apply where this happened, and those differ by country and change over time.
- Where the answer depends on a rule, a policy, or a deadline, say so plainly and tell me exactly who to ask and what to ask them.
- Work only from what I have told you. Do not invent policy names, reference numbers, deadlines, or amounts. If a detail matters and I have not given it to you, list it as something I need to find out.
Produce five sections, in this order:
1. ORDER OF OPERATIONS — who I should contact first, second, and third, and why that sequence rather than another. Note anything that would be premature to do yet.
2. EVIDENCE CHECKLIST — what to attach at each step, drawn from my list. Flag anything I have that is weaker than it looks, and anything obviously missing that I should try to obtain now.
3. THE ASK — one short paragraph I can paste, stating what happened and what I am requesting. Factual, unemotional, specific about the amount and the date. No threats, no adjectives.
4. QUESTIONS TO CONFIRM — the exact questions I should put to the airline, hotel, platform, card issuer, or relevant consumer authority to establish what I am actually entitled to and by when. Say which body answers which question.
5. IF NOTHING HAPPENS — what to do after no reply, what to log while waiting, and what my next escalation step would be.
Keep every section short enough that I can act on it today.
How the AI reads this prompt
Practical examples from different industries
A delayed flight that cost a hotel night.
A traveller's evening flight is delayed past the last connection, and they pay for an unplanned airport hotel plus a rebooked leg. Their instinct is to write an angry email about the four hours on the tarmac. The prompt's output instead sequences it: airline first with the specific out-of-pocket costs, receipts attached, before any card claim, because the carrier is the primary route and a card dispute filed first can complicate it. The QUESTIONS TO CONFIRM section gives them the exact things to ask the carrier about the delay's recorded cause and the applicable claim window — the two facts that determine everything, and the two the model refuses to guess at.
An undisclosed resort fee.
A couple checks out and finds a per-night fee that was nowhere on the booking page they saved. The evidence checklist immediately flags what matters: the saved booking confirmation and the original listing screenshot are the case, and the folio alone is not. The ASK is three sentences naming the amount, the dates, and the request for removal. The order of operations puts the hotel first and the booking platform second, since the platform's own listing is the discrepancy — and the questions to confirm are directed at the platform's fee-disclosure policy rather than answered by the model.
A charge that does not match the agreement.
A family returns a rental car and is billed for a fuel service charge despite returning it full, with a dated photo of the gauge. Here the prompt's value is restraint: the sequence is rental company first with the photo and the timestamped receipt, card issuer only if that fails, and the log of dates and reference numbers kept from the first contact onward. The IF NOTHING HAPPENS section is the part that actually gets this resolved, because the failure mode with rental disputes is not a rejection — it is silence, and knowing in advance what silence means is what keeps the claim moving.
Creative use case ideas
- The contractor who never came back. Half a bathroom, a deposit paid, and a phone that stops being answered. The structure transfers almost unchanged: sequence, evidence, a factual ask, and the questions that determine which body you escalate to.
- The gym membership that would not cancel. Recurring charges after a cancellation you are certain you made. The evidence checklist is the whole game here, and the prompt is very good at pointing out that "I cancelled in the app" is a claim, while a confirmation email is evidence.
- The community group and the venue deposit. A local club whose booking deposit was not returned after a hall was left clean. The unemotional ask matters enormously in small-community disputes, where the relationship survives the money.
- The mis-sold event ticket. A seat with an obstructed view that the seating map did not show. Screenshots of the purchase flow are the evidence, and the prompt will tell you to go and capture them before the listing changes.
- A student and a housing deposit. Deductions listed vaguely at the end of a lease. The same sequence-evidence-ask structure applies, and the "questions to confirm" step correctly routes to a local tenancy body rather than to a model's guess about local rules.
Adaptability tips
For a dispute involving several separate problems — a delay and a lost bag and a fee — run the prompt once per issue rather than bundling them. Combined claims dilute; each one has a different counterparty, a different evidence set, and often a different clock. The prompt says "one travel dispute" for exactly this reason, and it is worth honouring.
If you are outside your home country, add a line naming both jurisdictions and asking the model to note where the two might lead to different processes. It will not tell you which rules apply — it should not — but flagging that the question exists is genuinely useful, and it stops you assuming your home country's process travels with you.
To adapt this for a business expense dispute, swap "what I want" for the reimbursement outcome and add your finance team as a party in the order of operations. The sequencing question becomes internal as well as external, and the model handles that well if you name the internal step as a party.
If you want a gentler first contact, add "the first message should assume good faith and an administrative error" to THE ASK. Many charges are exactly that, and an opening that offers the other side an easy correction resolves a surprising number of these before anyone uses the word dispute.
Pro tips
Ask for THE ASK twice, once at 80 words and once at 200, and send the short one first. Long opening messages get skimmed and routed; short ones get answered. Keep the long version for the escalation, where detail actually helps.
Paste the model's QUESTIONS TO CONFIRM directly into your first contact as a short numbered list. Asking a company to state its own policy in writing is both useful information and a quietly effective move, because a written answer is a thing you can hold them to.
Start a dated log the moment you begin, and paste it back into the conversation before each new step. It keeps the model's advice consistent with what has already been tried, and the log itself becomes the evidence trail if this goes further than you expect.
Before you send anything, read the message once and delete every sentence that describes how you felt. It is nearly always two or three, and their removal makes the rest land harder.
Prerequisites
You need the specifics: dates, amounts, the name of the counterparty, and how you paid. You need your evidence gathered in one place, even if it is just a folder of screenshots. Most importantly, you need to know what outcome you actually want, stated as a number or an action — "I want the $180 resort fee removed" rather than "I want them to make this right." The prompt will ask you for that, and vagueness there produces a vague message that gets a vague reply.
Required tools
A general-purpose AI chat tool is enough. A tool that lets you save the conversation is a meaningful advantage, since disputes unfold over weeks and you will want to return to the same thread with new developments. If your tool supports document upload, having your booking confirmation to hand helps — but paste the relevant text rather than relying on the model to read a photo of a receipt accurately.
Frequently asked questions
Why will the prompt not tell me what I am owed?
Because it cannot know, and a confident wrong answer here is expensive. Compensation rules depend on where you flew, which carrier, what caused the disruption, what your fare conditions say, and what the rules are on the date in question — and they change. A model that recites a figure from its training data may be quoting a different country's regime entirely. The prompt is built to send you to the airline, card issuer, or consumer authority for that, and to make sure you ask them the right question.
Is going through the airline or hotel first really better than a card chargeback?
Often, though it depends on your card agreement and the nature of the problem, which is precisely why the prompt tells you to confirm rather than assume. As a general matter, card disputes are usually framed as a route when the merchant has failed to resolve things, and filing one prematurely can complicate a direct claim. Ask your issuer what their process expects before you file, and ask the merchant for a written response you can escalate with.
What if I have almost no evidence?
Then the evidence checklist becomes a to-do list rather than an inventory, and the useful move is to gather what still exists today. Bank statements, calendar entries, messages to friends sent at the time, and photos with timestamps are all more persuasive than memory. Say plainly in the prompt what you have and have not got — the output is much more useful when it knows it is working with a thin file.
Can I just have the AI write the whole complaint and send it?
You can, and you should read every line first, particularly any sentence that asserts a rule or an entitlement. If a sentence in the draft states what a company is required to do, either delete it or replace it with a question. A message that asks a company to confirm its policy is difficult to dismiss; a message that misstates the policy is easy to dismiss.
Recommended follow-up prompts
Try a "response triage" prompt: paste the company's reply and ask the model to identify what was answered, what was dodged, and what a one-paragraph follow-up should ask next. Dodged questions are where these disputes are won.
Pair this with the Week 6 protection-and-coverage prompt from earlier in the series. If you documented what your booking, card, or policy actually covered before you left, that document is the single most useful thing to paste into this conversation.
A "next-time prevention" prompt is a natural closer: ask the model what, given this specific dispute, you should capture at booking time in future — which screenshots, which confirmations, which terms saved as PDFs. That output belongs in the playbook you build below.
Tags and categories
Tags:
travel disputes, chargebacks, delay claims, consumer rights, evidence gathering, complaint letters, structured prompting, role constraints, intermediate prompts
Categories:
Travel Planning, Consumer Advocacy
Citations
The following are the authorities this variation directs readers to consult for their own case. They are listed as verification destinations, not as support for any claim made in this post about what a reader is entitled to.
- U.S. Department of Transportation, Aviation Consumer Protection (transportation.gov/airconsumer) — the U.S. authority for air travel complaints and passenger protections.
- Consumer Financial Protection Bureau (consumerfinance.gov) — U.S. guidance on disputing credit card charges and billing errors.
- Regulation (EC) No 261/2004, as published in EUR-Lex (eur-lex.europa.eu) — the EU air passenger rights regulation, for flights within its scope.
The Reusable Trip Playbook
Compress an entire trip into a playbook you can rerun.
Here is what usually happens to everything you learned on this trip. For about two weeks you remember it vividly — the booking site that was cheaper, the neighbourhood you would stay in next time, the mistake with the early flight. Then it fades, and eleven months later you open a blank browser tab and start from nothing, making a version of the same three mistakes with a different set of place names. This is not a memory problem. It is a storage problem: the knowledge existed, and there was nowhere to put it. A playbook is somewhere to put it. And the difference between a traveller who plans well and one who does not is rarely taste or experience — it is whether their process is written down anywhere or lives entirely in their head, being reconstructed from scratch each time.
Why this matters now
This week is the only week this prompt works properly. It depends on you remembering not just what you decided but why, and which decisions felt right at the time and turned out badly — a distinction that is gone within a month. If you have followed this series, you have also just lived through a full planning cycle with the reasoning made explicit at every stage, which is unusually good raw material. Run it now and the output is a personal system with your real numbers and your real preferences in it. Run it in October and you will produce a generic checklist that could have come from anywhere, which is worth roughly nothing.
You are going to help me build a reusable trip-planning playbook out of a trip I have just finished. Work in four phases and stop for my input between each one. Do not skip ahead to a later phase, even if you think you have enough.
PHASE 1 — INTERVIEW.
Ask me up to twelve questions, no more than four at a time, to establish: how this trip was planned and in what order; which decisions I made early and which I left late; which decisions I would make the same way again; which ones cost me money, time, or energy; and what I actually value when I travel, judged by what I chose rather than by what I say I prefer. If an answer is vague, push back once and ask for a specific example. Do not begin drafting anything until I tell you the interview is finished.
PHASE 2 — EXTRACT.
From my answers only, produce:
(a) My planning principles — no more than seven, each one sentence, each traceable to something I actually told you.
(b) My constraints — how I behave with money, what pace I can sustain, my tolerance for early starts and long transit days, and my genuine deal-breakers.
(c) My failure points — where this trip went wrong, and for each one, the earliest moment the problem was visible.
Mark clearly anything you inferred rather than heard, so I can correct it. Then stop and wait for my corrections.
PHASE 3 — BUILD.
Produce a planning playbook I can save and re-run for any destination. It must contain:
- A sequence of stages from first idea to booked trip. For each stage, state what gets decided there and what must NOT be decided yet.
- A prompt I can paste at each stage, written out in full and self-contained, so it works months from now with no memory of this conversation.
- My default assumptions, with the real numbers from this trip attached to each one, and a clearly marked blank wherever I have not given you a number.
- A decision checklist covering the two or three points where my failure points show I reliably get it wrong.
PHASE 4 — STRESS TEST.
Now try to break what you built. Name the three situations most likely to defeat this playbook — a destination unlike this one, a trip with different companions, a far shorter planning window — and for each, say specifically which stage fails and why. Then revise the playbook so it survives them. Finish by giving it a version number and a one-line changelog entry I can keep updating by hand.
Constraints throughout: build only from what I tell you. Do not add typical costs, generic itineraries, sample destinations, or assumptions about what travellers usually want. If you need a number or a fact I have not given you, leave a clearly marked blank rather than filling it.
How the AI reads this prompt
Practical examples from different industries
A family that takes the same kind of trip every summer.
Their interview surfaces that the recurring failure is not budget but sequencing: they pick dates around one parent's work calendar, then discover the good rental options went months earlier. The playbook's stage one becomes "lock the window before the reason," with an explicit note that accommodation research must not begin until dates are fixed but must begin the same week they are. That single ordering change, written down with a paste-ready prompt attached, is worth more to them than any amount of destination research.
A couple who split the planning unevenly.
One of them does all of it and resents it quietly; the other has opinions that arrive too late to act on. The extract phase makes this visible as a failure point with an early warning sign — the moment where opinions start arriving after bookings. The resulting playbook assigns stages rather than tasks, with a mandatory joint decision point before anything non-refundable is booked. It is a relationship fix disguised as a planning document, and it holds up better than a conversation about it did.
A traveller who mixes work trips with personal time.
The interview reveals two genuinely different planning modes that have been sharing one bad process — work travel is date-locked and expensable, personal travel is flexible and budget-constrained. The stress-test phase catches this directly, since "a trip with different companions" surfaces the same fault line. The revised playbook forks at stage one into two branches that share their booking and protection stages but nothing else, which is a structure they would not have thought to build unprompted.
Creative use case ideas
- The house move. Same shape entirely: a long sequence of decisions with hard ordering constraints, a budget that always runs over, and a set of lessons that evaporate before the next move. Run the four phases on a move you have just completed and you will produce something genuinely valuable to your future self.
- The annual holiday hosting playbook. Whoever cooks for twelve people every December has an enormous amount of undocumented process in their head. Extract it once and it becomes shareable, delegable, and survivable.
- The wedding or big-event planning system. Best built by someone who has just finished one, and enormously valuable to the next person in the family. The failure-point extraction is the part that saves them.
- The job search process. Stages, premature decisions, revealed preferences versus stated ones, and a documented set of your own hard-won rules. The structure of this prompt maps onto it almost without modification.
- The band tour or hobby-league season. Any repeating logistical undertaking with a fixed sequence and recurring mistakes. The version number matters most here, because these genuinely do improve over several iterations.
- The semester or study plan. A student who has just finished a hard term knows exactly where the process broke and will have forgotten by September. Four phases now, better term later.
Adaptability tips
If your trip was not planned in the structured way this series describes, run it anyway and let the interview do more work. The prompt does not assume you followed a method — it is designed to reconstruct whatever process you actually used, including a bad one. A playbook built from an admittedly chaotic trip is still a considerable improvement on nothing.
To build a household playbook rather than a personal one, run the interview with both travellers present and answering, and add a line asking the model to note where your answers conflict. Those conflicts are the most useful output in the entire exercise, and they are invisible when only one person is interviewed.
For a shorter version, collapse phases 2 and 3 and skip the stress test. You will get something usable in twenty minutes rather than an hour. Keep phase 1 intact regardless — the interview is where all the substance comes from, and a playbook built on four hasty answers will read like it.
To extend it across trips, re-run phase 4 alone after your next trip, feeding in the existing playbook plus what happened. That is what the version number is for, and a playbook on its third revision is a meaningfully different instrument from one on its first.
Pro tips
Before you start, paste in the outputs from Variation 1 and Variation 2 — your reconciliation assumptions and anything you learned from a dispute. The playbook's default assumptions section becomes dramatically better when it has real numbers to attach, and a dispute is a failure point with unusually good documentation.
When the model asks its interview questions, answer in full sentences with specifics rather than in fragments. The extract phase can only be as good as the transcript, and this is the one place where typing more genuinely produces a better result.
Save the finished playbook somewhere you will actually find it in eleven months, and put the words "trip planning" in the filename. More of these documents are lost to filing than to quality.
When you next use the playbook, keep a note of every place you deviated from it. Those deviations are the changelog entry for the next version, and they are far more honest than trying to remember what worked.
Prerequisites
You need roughly an hour and a willingness to answer questions honestly, including the ones about what went wrong. It helps enormously to have the reconciliation from Variation 1 done first, since the numbers give the playbook something concrete to hold. If you followed the series, having your earlier outputs to hand — the budget work, the booking decisions, the in-trip notes — makes the interview phase much richer, though the prompt is built to work for someone who has only just returned and read nothing before this.
Required tools
A general-purpose AI chat tool that supports a long multi-turn conversation. This prompt runs over four exchanges and produces a substantial final artifact, so a tool with generous context handling gives better results than one that starts forgetting the interview by phase 3. Free tiers can handle it, though you may want to run phases in a single sitting rather than across days. Somewhere to save Markdown or plain text — a notes app, a document, a repository — matters more than which AI you use.
Frequently asked questions
Will the AI actually stop between phases, or will it write everything at once?
Most current models respect the instruction, but not all of them, and some will drift back into completing everything after a long exchange. If it races ahead, reply with "stop — you are still in Phase 1, ask me the remaining questions" and it will return to the sequence. Restating the phase you are in is more effective than restating the rule, because it re-anchors the model's sense of where it is.
What if my trip was a disaster and I do not want to relive it?
A difficult trip produces a better playbook than a smooth one, because the failure points are unambiguous and you remember the warning signs vividly. You can also narrow the interview by telling the model which subjects to leave alone — it will work around a gap if you name it. The purpose is not to relitigate the trip; it is to make sure the next one is different.
How is this different from downloading a trip planning template?
A downloaded template encodes someone else's preferences, constraints, and mistakes, which is why most of them go unused after the first attempt. This one is built entirely from your answers, contains your actual numbers, and specifically targets the two or three places where you personally get it wrong. The structure is generic; the content is not, and the content is what makes a template survive contact with real use.
Do I have to have followed the whole series for this to work?
No. The prompt reconstructs your process through the interview, so it works for anyone who has just returned from a trip they can remember. It is genuinely richer if you have eight weeks of documented decisions to draw on — there is more to distil, and the reasoning behind each choice was made explicit at the time. But a traveller arriving here cold with a fresh memory and a pile of receipts will still get a working playbook.
What do I do with the playbook once it exists?
Save it, then open it as the first act of planning your next trip rather than opening a search engine. Run its stage-one prompt as written, and note anywhere it does not fit the new destination. Those notes become version 1.1, and by version 1.3 you will have something that plans a trip in a fraction of the time this one took.
Recommended follow-up prompts
Run a "playbook dry run" prompt on a destination you are only idly considering: paste the finished playbook and ask the model to walk stage one exactly as written, then report which instructions were ambiguous. Ambiguity is much cheaper to find on a trip you are not taking.
Try a "preference drift check" a year from now — paste the playbook alongside what you actually did on the next trip, and ask where your behaviour diverged from your stated principles. That is the honest input for version 2.
Pair the playbook with the Week 1 feasibility prompt for your next destination, but run the playbook first. If the two disagree about where to start, the disagreement itself tells you something worth writing into the changelog.
Tags and categories
Tags:
reusable prompts, planning systems, multi-phase prompting, personal frameworks, trip planning template, self-critique prompting, revealed preferences, advanced prompts, versioning
Categories:
Travel Planning, Prompt Engineering
Citations
NOT APPLICABLE
Which of the three should you use?
These three are not a ladder, and reading them as beginner-to-expert would be a mistake this week in particular. They are three separate jobs that happen to arrive at the same moment. The Honest Ledger is arithmetic with a memory attached: it takes fifteen minutes, it needs nothing but your statement, and its output is three sentences you will reuse. The Calm Claim is a process artifact for a specific problem — if no charge is wrong and nothing went badly enough to pursue, you can skip it entirely without losing anything. The Reusable Trip Playbook is the only one of the three that is not about this trip at all. It is about every trip after it.
The overlap sits mostly between the first and third. The reconciliation produces the real numbers that make the playbook's default assumptions worth having, and the failure points the playbook extracts are frequently the same misses the ledger surfaced in dollar terms. If you have time for two, do those two, in that order. The dispute prompt stands apart because it is triggered by circumstance rather than by the calendar, though it feeds the playbook a useful lesson about what to capture at booking time — which is why it earns its place in a week about closing things out well.
Choose by what is in front of you. If the statement is the thing bothering you, start with Variation 1 and stop there if that settles it. If something specific went wrong and you have been avoiding dealing with it, Variation 2 is the one that turns avoidance into a task with a next step. If you have finished the series and want the thing that makes all of it compound, Variation 3 is the whole point — it is the prompt that turns eight weeks of individual answers into a method you own, which is a fair description of what this site has been arguing for from the beginning.
TAGS:
