AI Guides › Everyday Life
By Nigel Guy · 3 min read
Grief and serious illness generate an enormous amount of genuine administration — hospital paperwork, a funeral to organise, people to notify, a household to keep running while everything else has stopped. It's tempting, and not unreasonable, to hand as much of that as possible to AI, because the logistics are real and exhausting and the tool is genuinely good at them. The mistake is letting that competence at the logistics quietly suggest it has something useful to say about the rest of it, which it doesn't.
The rule: AI can take real weight off the administration around grief and illness, and it has nothing to offer the part that isn't administration — it can't grieve for you, it can't tell you how you should feel, and treating its fluency on the subject as insight is where this goes wrong.
| Genuinely a task (AI can help) | Not a task (it can't) |
|---|---|
| Drafting a notification to send to a long list of people | Deciding how you feel about who to tell first |
| Organising paperwork, checklists, and deadlines around a death or diagnosis | Making sense of the loss itself |
| A first-pass structure for a eulogy or a difficult letter, for you to make actually yours | The words that need to come from you, however imperfect |
| Explaining an unfamiliar medical or legal term in plain language | Deciding what a diagnosis means for how you want to spend the time you have |
| Keeping a shared household running — meals, bills, the logistics guide covers this directly | The grieving, the sitting with it, the parts that don't resolve into a checklist |
AI answers fluently on almost anything, including grief, because it's been trained on an enormous amount of writing about it — which means it can produce something that sounds wise, comforting, or insightful about loss without actually knowing anything true about your loss. That fluency is the risk. A grieving person, already looking for anything that helps, can read generated comfort as a kind of answer, when what's actually happened is a plausible arrangement of words about grief in general, not a response to yours.
The same applies to illness. AI can organise information about a condition clearly. It cannot tell you what your specific prognosis means, what decision is right for your specific situation, or how to feel about any of it — and a confident-sounding answer here can do real harm precisely because it sounds like expertise (see the health-scare guide for the same caution applied to a shorter, more acute moment).
Skip asking AI what to feel, or whether what you're feeling is normal — that question deserves an actual person: a friend, a grief counsellor, a support group of people who've been through something similar, not a general-purpose model's synthesis of what grief writing tends to say. Skip using generated text as the final version of anything meant to be heard as your own words at a funeral or in a difficult letter — draft with it, then make it actually yours, the same discipline as the hard-conversation guide. Skip treating a fluent, sympathetic-sounding response about a diagnosis as medical guidance; it isn't, however carefully worded.