← All Articles
AI for Real Life

The hard email, made easy: apologies, complaints, and big asks.

The complaint, the apology, the raise request — they sit in drafts for days because the emotional load is real. Brain-dump the messy truth; let AI shape it into something sendable.

September 29, 2026 · 5 min read

Everyone has a drafts folder with a graveyard in it: the furious complaint never sent, the apology that felt impossible to word, the ask that felt too presumptuous. The blocker is never typing — it's emotional calibration. Too angry and you burn a bridge; too soft and nothing changes. AI turns out to be an excellent calibration instrument, because it has no ego in the fight.

The method: brain dump, then shape

Don't try to write the email. Write the truth — messy, unfiltered, exactly as you'd vent to a friend. Then hand it over with the shaping instructions. The two-step separation is the whole technique: you supply honesty, AI supplies diplomacy.

The universal shaper
Turn my notes below into a [firm but polite complaint / sincere apology / confident request] email. Keep all the facts, remove the emotion, keep it under 200 words. Tone: [professional / warm / direct]. End with a clear next step.

My notes:
[vent freely — the angrier/messier the better, nobody sees this part]

Now the three situations where this earns its keep:

1. The complaint that gets results

Effective complaints have a structure: what happened (facts, dates), what it cost you, what you want. Most failed complaints skip the third part — they vent without asking. AI enforces the structure. Try the shaper with "firm but polite complaint" and add: "Ask specifically for [refund / replacement / escalation to a manager]."

2. The apology that actually lands

Bad apologies explain, excuse, and subtly blame ("I'm sorry you feel that way"). Good apologies do three things: name what you did, acknowledge the impact, say what changes. Feed your messy guilt into the shaper with "sincere apology":

Apology calibration
Turn this into a sincere apology. It must: name specifically what I did wrong, acknowledge how it affected them (no excuses, no 'but'), and say what I'll do differently. Under 150 words. Warm, not groveling.

"Warm, not groveling" is the line that saves it — uncalibrated AI apologies tend toward either corporate or melodramatic. One adjective fixes the register.

3. The big ask: raises, fees, and favours

Asks fail from vagueness ("I was wondering if maybe sometime...") or from burying the number. The fix is structural: context, evidence, the specific ask, easy yes:

Ask calibration
Turn my notes into a confident, concise request email. Structure: one line of context, 2-3 bullet points of evidence/reasons, the specific ask with the number, and a low-pressure close that makes saying yes easy. Professional and direct — no hedging language like 'just' or 'sorry to bother you'.

That last instruction — stripping "just" and "sorry to bother you" — is small and transformative. Hedging language teaches the reader to discount you before they've finished the first line.

Before you hit send: the 3-check ritual.

1. Read it in their shoes. Ask the AI: "read this as the recipient — what would they feel?" It catches accidental sharpness you'd never see yourself.

2. Sleep on the angry ones. Draft tonight, send tomorrow. The email will still be there; the adrenaline won't.

3. Never paste anything you couldn't defend. Assume every email gets forwarded. AI-shaped doesn't mean AI-decided — the judgment call to send is always yours.

The deeper lesson: writing hard emails was never a writing problem. It was an emotional regulation problem wearing a writing costume. Offload the regulation — the tone calibration, the structure, the second pair of eyes — and what's left is just the decision to send. That part was always yours anyway.

Free newsletter

Get guides like this in your inbox

Practical AI for parents, teachers, and everyday people. Copy-paste ready. Always free.