Interactive exercise
Eighteen tasks from a working week and a home life. Choose who should do each one, AI, AI with you deciding, or you alone, and see our call with the reason.
No sign-up. Nothing you choose is saved or sent anywhere.
House rule: “Hand it to AI” means AI does the job and you glance at the result. In this game it never sends, pays, books, deletes or shares on your behalf without you, and anything it produces stays reversible.
Every card, with our call behind each one.
Work
Our call: AI drafts, you decideYou decide
AI can pull the numbers, draft charts and propose an outline, but the story the board hears and the sign-off on every figure are yours to defend.
Common mistake: Letting AI choose the narrative, or turning down its help with the data pulls and slide formatting that eat the evening.
Life
Our call: Hand it to AI
It is copying dates from one format to another, and the preview is where you catch a slip.
Common mistake: Importing without the preview, or feeding it a calendar from the wrong year or grade.
Work
Our call: You, AI stays outYou decide
Preparing with HR, and even rehearsing, is preparation; the conversation itself, and the accountability for it, has to be yours.
Common mistake: Confusing the two, and letting a prepared script stand in for listening to the person in front of you.
Work
Our call: Hand it to AI
You were in the room and the summary is only for you, so a quick read of the result is check enough.
Common mistake: Forwarding the summary to others without reading it, or using this shortcut on a meeting that held confidential decisions.
Life
Our call: AI drafts, you decideYou decide
AI can explain the terms and draft your questions; what you dispute and what you pay is your decision with the insurer and provider, and you leave out identifiers it does not need.
Common mistake: Pasting the full bill with account numbers into a tool, or treating its reading as the final word.
Work
Our call: You, AI stays outYou decide
A hiring decision carries accountability, fairness obligations and a working relationship that no tool can own for you.
Common mistake: Letting an AI ranking settle it because it feels objective.
Work
Our call: Hand it to AI
Applying a style you have defined is mechanical, and you can spot-check it in a minute because nothing is removed.
Common mistake: Letting the tool “clean up” by guessing missing details or merging people who only look like duplicates.
Life
Our call: AI drafts, you decideYou decide
AI saves you the evening, but you know the allergies, the budget and who needs a lift, so you review before anything goes out.
Common mistake: Sending the invites and buying the food straight from a generic plan.
Work
Our call: You, AI stays outYou decide
On the call, repair depends on them hearing you take responsibility and respond to what they actually say.
Common mistake: Skipping the call and sending a written reply, however well drafted, when the moment needs a voice. The follow-up email afterwards is a fair place for AI to draft and you to decide.
Work
Our call: AI drafts, you decideYou decide
AI can organise your own notes and examples into a clear structure, but the judgment and the rating are yours, and personal details belong only in tools your employer has approved.
Common mistake: Asking AI to “write a review” from a name and an impression, which produces confident text with no evidence behind it.
Life
Our call: Hand it to AI
Grouping by date and place is what these tools do well, and suggestions you can undo cost nothing.
Common mistake: Letting a tool reorganise or “clean up” the originals, or using one that uploads family photos somewhere you did not intend.
Work
Our call: AI drafts, you decideYou decide
AI is fast at a first map of the players and questions, but it can be confidently wrong, so you verify the claims you will repeat.
Common mistake: Quoting an AI summary in a meeting without checking a single source.
Work
Our call: Hand it to AI
Once the rules are explicit, finding open slots is mechanical, and you still choose which to send.
Common mistake: Expecting it to know unwritten preferences, like the colleague who never takes Friday afternoons.
Life
Our call: AI drafts, you decideYou decide
AI can find repeat charges quickly, but not every repeat charge is a subscription, and cancelling is your call.
Common mistake: Connecting a tool to your whole account when a trimmed, exported statement would do.
Work
Our call: Hand it to AI
Reading a receipt and filling fields is mechanical, and submitting stays a separate step you take.
Common mistake: Assuming an approval step downstream will catch a wrong amount or currency.
Work
Our call: AI drafts, you decideYou decide
AI gives you a fast structure, but you decide what the role really needs and check the wording for requirements that quietly exclude good people.
Common mistake: Posting the draft as is, inflated requirements and all.
Life
Our call: You, AI stays outYou decide
What repairs a friendship is you owning it, listening, and following through; that happens between the two of you.
Common mistake: Putting the effort into perfect wording and none into the conversation and what you do next.
Life
Our call: AI drafts, you decideYou decide
AI interviews you with questions about him. You answer out loud, in whatever order the memories come. AI shapes what you said into a rough draft, and you edit it into your own words.
The memories and the voice are yours; AI helps you get them out and into order.
Common mistake: Asking for “a moving eulogy” with nothing of him in it, and getting something that could be about anyone.
Keep reading
Article
People reach for a chatbot by reflex, but language models are one small family of AI. What the specialists do, when to use each, and a tool to help you pick.
Article
The most sought-after role in AI, why the big labs have committed around nine billion dollars to hiring them, and what it would mean to have one for a single person instead of a company.
Guide
The short answer is nothing. The longer answer is four things worth having open, and what you walk out with.
Interactive tool
Answer two questions and get the right kind of model for your problem, in plain terms and in technical detail. Includes a map of 145 approaches, because language models are one family out of fourteen.
