← All workflows

Workflow

Build your Email Prompt Bank: five prompts that handle 80% of your inbox (Day 8 of the 30-Day Challenge)

✓ TestedFounderFor Founder & Operator
Time saved30–45 minutes of inbox time a day

The task

Build the five email prompts you'll reuse daily — reply, follow-up, summary, tone-shift, hard conversation — and save them as your Email Prompt Bank. This is the Day 8 build from the 30-Day AI-Native Challenge, fully worked.

Before AI

Email eats the morning: rereading threads to remember context, drafting replies you rewrite twice, softening a blunt message for ten minutes. None of that is judgment work — it's formatting.

What you'll need

  • Any chat model (free tiers fine)
  • Your Prompt Library doc from Day 2
  • Three real threads from your inbox to test on (sanitized if you're on a consumer tier)

The workflow

1. The reply prompt. The workhorse. Save this with your own voice notes filled in:

Code
You draft email replies in my voice: [direct but friendly / brief / no exclamation
points / sign-offs are just my first name].

Here's the thread:
[paste thread]

Draft a reply that: [accepts the meeting but moves it to next week / declines
politely and offers an alternative / answers their two questions].

Under 120 words. If the thread contains a question I haven't given you an answer
for, put [NEED FROM ME: ...] instead of inventing one.

That last line is the discernment habit from Day 4 — the prompt refuses to hallucinate on your behalf.

2. The follow-up prompt.

Code
Draft a follow-up to this thread. It's been [X days] with no reply.

Tone: assume good faith — busy, not ignoring me. One sentence of context re-anchor,
one clear ask, one easy out ("if this isn't your area, who should I ask?").
Under 75 words. Do not open with "Just following up."

3. The thread summary prompt. For the 40-message chain someone forwards you:

Code
Summarize this thread for someone who just got added:
1. What's being decided
2. Current positions, by person
3. What's blocking
4. The one thing I need to do, if anything
Four short sections. If nothing needs my action, say so explicitly.

4. The tone-shift prompt. For when your first draft is honest but career-limiting:

Code
Rewrite my draft below. Keep every substantive point. Remove the frustration.
Target tone: professional but human — firm on the facts, generous to the people.
Flag any sentence where the rewrite changed the meaning, not just the tone.

[paste draft]

5. The hard-conversation prompt. Escalations, pushback, saying no to a client:

Code
Help me write a hard email. Situation: [what happened]. My goal: [what I need to
happen]. What I must preserve: [the relationship / the contract / my boundary].

First: give me 3 possible framings (direct / collaborative / formal), one sentence
each. I'll pick one, then you draft it.

Note this one is two-step on purpose — for high-stakes email, you want to choose the framing before any words exist to anchor you.

6. Save all five in your Prompt Bank with "use when" lines. Test each on a real thread from your inbox today.

Verify it worked

Tomorrow morning, run your actual inbox through the bank for 30 minutes. Count: how many replies went out with ≤1 edit? Above half, the bank works. Below that, the voice notes in prompt #1 need another pass — add three real emails you've written as style examples.

Troubleshooting

  • Replies don't sound like you? Paste 2–3 of your own sent emails into the reply prompt as voice examples. Style is shown, not described.
  • Model invents commitments ("happy to jump on a call Thursday!") — tighten the [NEED FROM ME] instruction; never let a draft promise what you didn't authorize.
  • Summaries miss the point? Add "optimize for what changed since I last read this thread."

Reality check

This bank handles the 80% that's formatting. The 20% that's actual judgment — negotiations, sensitive personnel notes, anything legal — gets the two-step treatment or no AI at all. Knowing the difference is the skill.

Data & security

Email threads are exactly where sensitive data leaks into consumer AI tools. Names, deal terms, personnel details: enterprise tier or sanitize. Your Day 6 Data-Boundary Playbook governs — if you haven't written it yet, that's the day to take seriously.

Going further

Day 22 of the challenge automates the next layer: incoming form/email → AI summary → draft → your approval. The first-automation workflow builds it end to end.

Your takeaway

Five saved prompts that turn inbox time into approval time — and the first proof that a prompt library beats prompting from scratch.

Source: Agentic Daily

Exact prompts included · Untested steps are marked · Corrections are public