How do I use AI for X?
You already know what you want to hand over. This is where to find it.
Every one of these is a walkthrough of a real, repeating piece of work, with the prompt written out — not advice about prompting in general. Pick the chore you did most recently and start there.
Find your task
Not listed? The full set is below — and if the job you keep repeating isn't here, it's a good candidate for building one yourself.
Roughly in order of effort: do the task once, make it repeatable, then hand the whole thing over.
Each of these takes one recurring chore and walks the whole thing, prompt included. Start with whichever you did most recently.
A step-by-step walkthrough for handing off a recurring report (the weekly summary, the monthly recap) to AI without losing the judgment that made it yours.
How to get AI action items from a meeting: a copy-the-prompt tutorial that turns a raw transcript into decisions, owners and due dates you can send out before the room has cooled.
AI cuts professional writing time by about 40%, but the study that found it used tasks with no real context and no relationships. Here's how to close that gap and get replies that actually sound like you wrote them.
The safest way to use AI on a spreadsheet is to never let it hand you a number. Ask for the formula instead: it's auditable, it recalculates, and it's the one thing decades of spreadsheet research says actually catches errors.
Turning a thing that worked into a thing you can rerun — which is the difference between a trick and a workflow.
Stop re-explaining yourself in every new chat. Where the feature lives in ChatGPT, Gemini, Claude and Copilot, what to write in it, and how to test it: for one job you do constantly.
A hands-on companion to the commands-vs-skills-vs-plugins guide: build one genuinely useful thing in Claude Code, starting at the simplest rung and climbing exactly one step when you feel the ceiling.
Three words that get thrown around whenever people customize Claude Code, and a plain-English guide to what each one is, its trade-offs, and when to reach for it.
Most people use Claude like a chatbot and never touch the four built-in capabilities that do the real work: file creation, Projects, Artifacts, and web search. A plain-English tour of what each does, when to reach for it, and where each one bites.
When the task spans several steps and several apps, and you would rather not be the one carrying things between them.
You don't build an agent, you configure one, and it comes down to four decisions, none of which involve programming. The fourth is the one nobody mentions and the one that keeps you out of trouble.
In one week this month, Anthropic put Claude Cowork on phones and OpenAI shipped ChatGPT Work. Both do multi-step jobs with little supervision. Here's how to pick the first task to hand over, and the one to keep.
Models score around 43% on simple character-counting tasks. Everyone blames tokenization. A larger benchmark found it barely predicts counting errors at all, and what does predict them is worse news for real work.
Every assistant now has a thinking mode. Research found three levels of difficulty, and on the easiest one the ordinary model beats the reasoning one. A plain-English guide to which tasks want it.
Anthropic sampled a million conversations and found experienced users delegate less, not more. What they do instead, and how it fits the long-chat research.
AI presentation tools give you clean slides and an empty argument. Why default AI slide decks fail, and the assertion-evidence fix that takes ten minutes.
Forty messages in and the AI ignores your instructions. It isn't memory. It's the context window, and accuracy collapses in the middle. Four habits that fix it.
A lab ran one identical prompt 1,000 times on the setting meant to remove all randomness. It produced 80 different answers. The cause isn't what most people assume, and it changes how you should build anything repeatable on top of AI.
Most people brace for made-up facts. But half of the AI errors managers actually catch are failures of context, and those look completely correct on the page. Here's what to check instead, and how long it should take.