🔁 5 AI Habits That Save Me A Day Every Week

The internet is drowning in AI hacks. “This one prompt will change your life.” And maybe you’ve tried a few - screenshotted a few, meant to try them, and the screenshots folder is where they retired. That’s the thing about hacks: they’re borrowed. They don’t survive contact with a busy Tuesday.

Habits are different. A habit doesn’t need remembering, motivation or a saved post - it just fires, the way you don’t decide to check the mirror before pulling out. After six years running the back ends of other people’s businesses, and a few thousand hours learning AI properly when it started circling my industry, these are the five that stuck. Not the cleverest tricks I know… the ones that survived.

Do they genuinely save me a day a week? Totted up honestly - drafts I didn’t start from blank, context I didn’t re-explain, mornings that started at speed - yes, somewhere around a working day, spread invisibly across the week. Your number might be smaller to start. It grows, because habits compound and hacks don’t.

Each habit below comes in three parts: what it is, how to start it, and the prompt that anchors it. A habit with a prompt attached is much easier to keep, because you’re never wondering what to type.

The Anchor Five

1. The ask-first reflex. What it is: before I google something, open a blank document, or sit staring at a task I’m dreading, I ask Claude first. Not because it always nails it - because asking first means every task starts with momentum instead of a sigh. How to start it: for one week, put a note where you’ll see it - “ask first” on a sticky, on your monitor, wherever - and every time you catch yourself about to slog, ask instead. By day four it starts firing on its own. The non-obvious bit: the reflex matters more than the answer. Even when Claude’s first pass is mediocre, you’re now reacting to something instead of generating from nothing, and reacting is a fraction of the effort. The habit pays even when the output doesn’t.

Nick this one - the anchor:

Here's the task in front of me: [DESCRIBE IT, roughly is fine]. Either give me a solid first pass at it, or - if you need more to do it well - ask me up to three questions first. Don't do both, and don't pad it.

2. Drafts, never blank pages. What it is: a house rule that I never open an empty document. Every report, email, plan or awkward message starts life as a Claude draft that’s allowed to be wrong. How to start it: make it binary, not aspirational - the rule is “no blank pages”, full stop, so there’s no daily decision to make. The non-obvious bit: a wrong draft is nearly as useful as a right one. The moment you read a draft and think “no, that’s not it at all”, you suddenly know what it actually is - being mildly annoyed at a draft is one of the fastest thinking tools there is. The draft’s job isn’t to be good. It’s to end the staring.

Nick this one - the anchor:

Rough first draft of [THE THING] - for [WHO IT'S FOR], roughly [LENGTH], sounding [TONE]. Don't polish it and don't hedge with three options. One committed draft I can tear apart.

3. The end-of-day wrap. What it is: two minutes at the close of the working day telling Claude what happened, what’s unfinished, and what tomorrow’s first move is. How to start it: staple it to something you already do without fail - just before you close the laptop, while the kettle’s on for the last brew. Two minutes, genuinely. The non-obvious bit: this isn’t journalling, it’s a handover to tomorrow-you - and tomorrow-you is the one who profits. The most expensive minutes of my day used to be the first twenty, spent reconstructing where I’d got to. Now the day starts with a note that says “start here”, written by someone who was there yesterday. With memory switched on, those wraps quietly become a running record of every project you’ve touched. (If yours isn’t on: Settings → Capabilities.)

Nick this one - the anchor:

End-of-day wrap. Today I: [WHAT GOT DONE]. Still hanging: [WHAT DIDN'T]. Rattling round my head: [ANY LOOSE WORRY]. Give me back: 1) tomorrow's first move, one line, 2) anything I've said that sounds like a deadline I should note, 3) permission to log off, briefly.

4. One chat per project. What it is: instead of forty scattered conversations, each ongoing piece of work - a client, the house move, the big report - lives in its own named chat (or a Project, once something deserves its own instructions and files). How to start it: takes ten minutes once. Name a chat per live project, and when you return to that work, return to that chat. The non-obvious bit: the win isn’t tidiness, it’s never re-explaining. Context is the tax you pay every time you start fresh - “so I run a small business, and there’s this client, and last month…” - and one-chat-per-project means you pay it once. Asking “given everything so far, what am I forgetting?” in a chat that has three weeks of context produces answers a fresh chat simply can’t.

Nick this one - the anchor (first message in any project chat):

This chat is home base for one project: [NAME IT]. The gist: [THREE LINES - what it is, where it's up to, what done looks like]. Keep all of it in mind whenever I come back here. First job: [TODAY'S TASK].

5. The Friday review. What it is: fifteen minutes at the end of the week - what moved, what stalled, and the three things that matter next week - with Claude as the reviewer who never forgets to show up. How to start it: calendar slot, Friday, recurring, before the weekend feeling kicks in. Lower the bar to “even five minutes counts” so it survives busy weeks. The non-obvious bit: the sneaky-valuable question isn’t “what did I do?” - it’s “what did I ask you repeatedly?”. Anything you’ve prompted three times this week is a system waiting to be built: a template, a Project, a standing instruction. The Friday review is where one-off saves get promoted into permanent ones. That’s the compounding engine of this whole guide, hiding in item five.

Nick this one - the anchor:

Friday review, 15 minutes. This week: [WINS, ROUGHLY]. Stalled or nagging: [THE REST]. Walk me through: 1) what actually moved forward, 2) what's stalled and the smallest unsticking step for each, 3) my top three for next week - one of them visible to someone other than me, 4) anything I asked you for repeatedly this week that we should turn into a template or standing instruction.

A worked example, so you can see it stack

Say you’re a bookkeeper with five clients. One chat per client (habit 4) means each chat holds that client’s quirks - the one who always pays late, the odd VAT arrangement. Month-end letters start as drafts, never blank pages (habit 2), inside the right chat, so they arrive pre-loaded with context. The day ends with a wrap (habit 3): “finished the Hendersons, stuck on a missing invoice for the café”. Friday’s review (habit 5) spots that you’ve asked for “chase email for missing receipts” three times… so it becomes a template, and next month that job takes thirty seconds. No single step saved more than twenty minutes. The stack is where the day comes from.

Honestly, the limits

Habits are slower than hacks - that’s the deal. Expect two or three weeks before any of these fire without the sticky note, and expect to skip days, because everyone does; a skipped day is a blip, not a failure, and the habit survives anything except giving up on it entirely. Also, fair warning: none of this reorganises your personality. I have ADHD, I still lose entire mornings to nonsense sometimes, and these habits don’t prevent that - they just mean the recovery is faster, because the wrap knows where I left off even when I don’t.

The one to nick first

The end-of-day wrap. Two minutes, easiest to remember, and it’s the one that makes tomorrow better rather than just today - which is exactly the shape of thing that keeps itself going.

Start with one habit, run it for a fortnight, then add the next. And if you’re curious what this looks like fully grown, What AI Actually Does In My Week is the same idea, zoomed out.