AI for travel, recipes, and life admin

The use cases that turn a curious first-time user into someone who reaches for ChatGPT every day are usually small. Not "transform my business" — just the kind of thing you used to do badly in a browser tab and feel slightly relieved when it was over.

This article is a tour of those wins. Real prompts, real situations, no career advice. Once a handful of these become reflexive, the habit takes care of itself.

Travel: turning "I should plan that trip" into an actual trip

Travel is where AI shines, because the work is mostly synthesis: weighing options, checking compatibility, building a sensible order. Try:

Help me plan a four-day trip to Lisbon in late September. Travelling: me and my partner, mid-30s. We like good food, walking, and small interesting museums. We hate crowds, fancy restaurants, and anything that involves more than two transfers a day. Budget for hotel: ~€150/night. We arrive Friday evening, leave Tuesday morning. We have already booked the flights.

>

Before drafting an itinerary:

>

1. Ask me five questions you would need answered to make this plan actually fit us. 2. Once I answer, give me a day-by-day plan with morning, lunch, afternoon, evening — naming specific places. Mark any place I should book ahead. 3. Add one alternative for each day in case of rain. 4. End with a "what to do if you have an extra two hours" list.

Notice the structure. You forced the model to interview you first, gave it constraints, asked for a concrete plan with backups. After thirty minutes you have something genuinely usable — better than the average travel blog, calibrated to your taste.

A useful follow-up: paste in your hotel reservation and ask "Given where I am staying, reorder the itinerary to minimize walking and transfers." The model is happy to do logistics; most travelers never ask.

For the verifying step: turn on web search before asking about specific restaurants and museums, especially recent openings or temporary closures. The model's training cut-off can be months old.

Recipes: cooking what is in your fridge

The classic "help me cook" prompt — "give me a recipe" — is useless because it ignores what you have. The version that earns its keep:

Here is what I have in my kitchen right now:

>

- In the fridge: half a leek, some sour cream, two eggs, a bag of spinach that needs using, half a chicken breast, parmesan, two carrots - In the cupboard: rice, pasta, tinned tomatoes, chickpeas, the usual oils and spices - I do not want to go to the shop

>

Give me three recipes I can make in under 30 minutes using mostly these ingredients. For each, list ingredients in the order you'll use them, the steps, and the trickiest part of getting it right.

>

Skip any recipe that needs a long list of things I don't have.

This produces three actually-cookable dinners. A common pattern is to follow up: "Make recipe 2 vegetarian," "Add a starch to recipe 1," "What wine would go with recipe 3?" — and you've solved your evening in ninety seconds.

For meal planning across a week: paste in your typical week, your dietary constraints, and your shopping habits, then ask for a "shopping list that produces five different dinners with maximum ingredient reuse." Same pattern, bigger scope.

Life admin: the small things that eat your weekends

Every adult has a backlog of small administrative tasks they keep meaning to do. Cancel that gym membership. Complain to that airline. Write to the city about the broken streetlight. Reply to that long email. AI is excellent at all of these because the hard part is not the words — it is overcoming the inertia of writing them.

Some templates that work:

Cancelling a subscription:

Write a short, polite, firm cancellation letter for [service]. I have been a customer for [X years]. They will probably try to retain me with a discount — I do not want to negotiate, just cancel. End with a clear request for written confirmation of the cancellation date.

Complaining about a service:

Write a complaint about [situation]. Tone: matter-of-fact, factual, not angry. List what happened with dates. State what I want as resolution. Keep it under 200 words. Avoid any sarcasm or threats — I want it to be the kind of complaint that a customer service manager will actually read and respond to.

Replying to a long-overdue email:

Here is an email I should have replied to two months ago. Draft a short reply that acknowledges the delay without grovelling, addresses the original question, and ends with a clear next step. Tone: warm, not apologetic. Under 100 words. Do not start with "I am so sorry for the late reply."

A small psychological note: many of these messages have been on your to-do list because you did not want to write them. Once you see the model produce a draft in ten seconds that is 90% of what you would have written, the inertia evaporates.

Plans, decisions, and one-page comparisons

The other category is the "I have to choose between X and Y" decision, where the work is laying out the comparison in a form you can scan. Try:

I am deciding between two pension fund products in Estonia: [Fund A] and [Fund B]. Build a side-by-side comparison table covering: fees, historical performance (with caveats), risk profile, what the fund actually invests in, and ease of switching providers later.

>

Then write a one-paragraph "if you are this kind of person, choose A; if you are this kind of person, choose B" decision guide.

>

Note: turn on web search and only use information you can attribute to the official fund documents or the Estonian financial regulator. Mark anything you are unsure about.

This is one of the highest-ROI uses of AI most people miss. It works for choosing pension funds, kitchen appliances, mortgages, software subscriptions, healthcare options — anything where the cost of choosing badly is real and the comparison is tedious to make.

The key is turning on search for anything that involves real-world specifics, and instructing the model to flag uncertainty rather than fill it in with plausible-sounding numbers.

The list that you can copy this week

If you want to make AI a habit, do these five things in the next seven days:

1. Plan one upcoming trip or weekend, using the structured prompt above.
2. Use AI to cook dinner with whatever is in your fridge tonight.
3. Send one piece of life admin you have been avoiding (a cancellation, a complaint, a chase).
4. Make one shopping list with ingredient reuse.
5. Compare two real options you have been putting off deciding between.

That is five uses, ten minutes each. By the end of the week you will have stopped thinking of AI as a thing for "important" tasks and started reaching for it on small ones — which is when the habit forms, and when it becomes hard to imagine a week without it.

The reason these uses build the habit is not that they are profound. It is that they are frequent, and each one ends with a small "huh, that was nice." That is the loop. Stack enough of them and you are someone who uses AI well.