Most people use AI wrong for a sympathy card

Today's Tip: Writing the Sympathy Card You've Been Putting Off

There is a blank sympathy card sitting on a kitchen counter somewhere in America right now. It has been sitting there for three weeks. And the longer it sits, the harder it gets to send.

Let me explain…

My friend Curtis is 70. He ran a used bookstore in Durham for 32 years before he retired. The man has written more book reviews and recommendation letters than I can count. Words are his thing.

And yet… when his old college roommate's wife passed away last month, Curtis sat at his kitchen table for four nights in a row, staring at a blank card. He told me he started six different drafts. He crossed them all out. One sounded too formal. One sounded too cheerful. One read like a Hallmark card. One mentioned his own grief too much.

By the fifth night, he was ready to give up and just send flowers. That's when his daughter mentioned she had been using Claude — the writing AI from Anthropic — to help her with hard work emails.

Curtis was skeptical. A computer writing a sympathy card felt cold to him. Wrong, even. But he was stuck… and he tried it.

Here's the part that surprised him. He didn't ask Claude to write the card for him. He asked Claude to help him write it himself. And the difference between those two things is everything.

He went to claude.ai, signed in with his Google account (free, no credit card), and typed in the prompt I'm about to give you. Claude didn't spit out a finished card. It asked Curtis three gentle questions about his friend's wife — what he remembered about her, what he wanted the family to know, and how close he was to them now. Then it gave him a draft.

Curtis changed three sentences. Added one of his own. Wrote it out on the card by hand. And it went in the mail the next morning.

Here's the exact prompt that did it…

What makes this prompt work isn't fancy language. It's the questions. By making the AI ask before it writes, you stop the machine from inventing memories you don't have… and you make sure your own voice is the foundation of the card.

Curtis said the hardest part — the part that had stopped him cold for four nights — was figuring out the order of the sentences. What goes first? What goes last? What do you put in the middle that doesn't sound hollow? Claude handled that structure problem in about 20 seconds. Curtis handled the heart of it. That's the right division of labor.

Now, I know what you're thinking. "Isn't this dishonest? If a computer helped me write it, isn't that fake?"

I understand the worry. But think about it this way. For decades, people have asked their spouses, their pastors, their friends — "Does this sound right? Should I say it this way or that way?" Getting help with the words has never been the same as faking the feeling. The card Curtis sent contained one specific memory he had of his roommate's wife from 1974, written in his own handwriting, sealed with his name. The fact that an AI helped him put the sentences in the right order doesn't change a single one of those things.

A blank card sitting on a counter for three weeks helps nobody. A card in the mailbox in three days does.

Try it the next time you face a hard letter. Sympathy card. Thank-you note after a long hospital stay. A wedding card for a second marriage where you don't know the new spouse well. A note to a grandchild's teacher you've never met. Anything where the blank page has been beating you.

Open claude.ai. Paste the prompt above. Answer the three questions. Send the card.

★ Quick Wins

Skill Builder: Using AI to Proofread and Edit

Here is a habit worth building. Whenever you write something that matters — a letter to your insurance company, an email to your doctor, a long note to a grandchild — run it past AI before you send it. Not to have it rewritten. Just to have it checked. This is one of the most useful things AI does, and almost nobody our age uses it this way.

The trick is to tell AI exactly what kind of help you want. "Proofread this" is too vague. The AI will start rewriting your sentences to sound more polished, which is usually not what you want. You want it to check your spelling, catch typos, point out a confusing sentence — and then leave the rest alone.

There is a difference between proofreading and editing, and you should know which one you're asking for. Proofreading is finding mistakes — typos, missing words, a "their" that should be "there." Editing is bigger — reordering paragraphs, sharpening unclear sentences, suggesting cuts. Most of the time, when you've written something heartfelt, you want proofreading only. The voice is yours. You earned it. Don't let the machine sand it down.

Here is the prompt I keep saved. I use it almost every week. You can save it in your phone's notes, a sticky on your desk, anywhere you can copy it from quickly.

One more thing worth knowing. AI is good at catching the small stuff and pretty good at catching confused sentences. But it cannot catch a fact that's wrong. If you wrote "the appointment is on Tuesday the 14th" and Tuesday is actually the 15th, the AI will not know that. Always check dates, names, addresses, and phone numbers yourself. The machine is your copy editor, not your fact checker.

Try this once today on something simple — a short email, a text to your kids, a note you've been meaning to send. Once you see how careful and quick it is, you'll start using it everywhere.