Workflow Examples — AI Workflows & Systems

You'll learn

Three complete workflow walk-throughs you can copy today
The shared skeleton behind every workflow: trigger, steps, stored prompt, review
How to adapt a borrowed workflow to your own context
Where the next frontier (connecting tools) fits — and why it waits
The capstone: your own process, fully documented

01Section

Three workflows, one shape

This lesson is the path put to work: three complete workflows, each laid out as trigger, steps, stored prompt, and review — the same skeleton you've been building all along. Copy whichever is closest to your life, run it once as-is, then adapt.

One honest note first: you won't find time-savings numbers here, because we'd be making them up. What these workflows reliably save is decisions — the task is already designed, so you just run it.

How to use these

Don't admire them — run one this week, as-is. Adaptation comes easier after one real run.

Checkpoint

02Section

Walk-through 1: Weekly content note → posts

For anyone who posts regularly and dreads the blank page. The trick: the input builds itself all week — you keep one running note of ideas, observations, and half-thoughts, and once a week it becomes drafts.

Trigger: Your content slot — say Friday morning, or whenever posts get drafted.
Gather the input: Open the week's running note. Don't clean it up — the prompt handles mess.
Run the stored prompt: Paste the note into the prompt below.
Review: Check facts, check it sounds like you, and cut anything the AI padded to fill space.
Deliver: Two or three drafts into your drafts folder or scheduler. Done.

Stored prompt — content note to post drafts

Below is my raw notes file of ideas and observations from this week. Pick the two or three strongest ideas and turn each into a short post draft in my voice: direct, plain language, no hashtags, no hype. For each draft, tell me which line in my notes it came from. If an idea is too thin to draft honestly, say so instead of padding it.

My notes:
[paste the week's note here]

Checkpoint

03Section

Walk-through 2: The inbox triage hour

For the backlog that builds up while you do real work. The workflow doesn't answer your email — it sorts it, so your attention lands on the few messages that deserve it.

Trigger: A fixed slot — daily at 4pm, or twice a week. The slot matters more than the frequency.
Gather the input: List the backlog as one line per message: sender, subject, what they're asking. Summarize — don't paste whole threads.
Run the stored prompt: Paste the list into the prompt below.
Review: Scan the sorting — especially anything marked low-priority. The AI doesn't know your politics; you do.
Work the list: Answer the "respond now" group, schedule the decisions, let the rest wait honestly.

Stored prompt — inbox triage

Below is my email backlog, one line per message (sender, subject, what they're asking). Sort every message into exactly one of four groups: RESPOND NOW (time-sensitive or blocking someone), NEEDS A DECISION (I have to think before replying), WAITING ON OTHERS (nothing for me to do yet), ARCHIVE (no action needed). Within RESPOND NOW, order by urgency and say in one line why. Don't draft any replies — just sort.

My backlog:
[paste your one-line-per-message list here]

Privacy note: Email is full of other people's information. Summarize senders and asks instead of pasting entire private threads, and leave out anything sensitive — triage works on the shape of a message, not every word of it.

Inbox Triage prompt

Checkpoint

04Section

Walk-through 3: Meeting notes → action list

The workflow this path has used as its running example — here it is complete. It turns the meeting you just left into the list your team actually needs.

Trigger: A meeting ends and you took notes.
Gather the input: Paste notes under a fixed header: meeting name, date, attendees, then the raw notes.
Run the stored prompt: The prompt below — note its rules against inventing owners and dates.
Review: Check every owner and date against your memory of the room. This is the step that keeps trust in the output.
Deliver: Post the list where your team works. Chase the UNASSIGNED rows — those are real gaps the meeting left.

Stored prompt — meeting notes to action list

Turn the meeting notes below into an action list. Format: a table with Task, Owner, Due date. Rules: only include actions actually stated in the notes; if no owner was named, write UNASSIGNED; if no date was given, write NONE — never invent either. After the table, separately list anything that sounded like a decision.

Meeting: [name]
Date: [date]
Attendees: [names]
Notes:
[paste raw notes here]

Quick knowledge check

All three walk-throughs share the same skeleton. What is it?

AA specific AI tool, a paid plan, and an integrationBA trigger, repeatable steps, a stored prompt, and a human reviewCA daily schedule and a shared team documentDAn input template, but the rest varies each run

Checkpoint

05Section

Copy, adapt, then make one yours

Adapting a borrowed workflow isn't mostly about rewriting the prompt — it's about refitting the input shape, the output format, and the review rules to your context. Your meetings have different politics; your inbox has different stakes; your voice isn't anyone else's. Change those parts first and the prompt usually needs only light edits.

And a last honest note about the frontier past this path: tools can be connected so steps trigger each other with less of you in the loop. That's real, and worth exploring eventually — but earn it. Run the human version until it's boring and the review step almost never catches anything. Automating a workflow you haven't made reliable just produces mistakes faster.

Pause and think: Which of the three walk-throughs is closest to something you already do? Start your adaptation there — familiarity makes the refitting obvious.

Build With AI

Checkpoint

Prompt exercise

Capstone: document your one repeatable process

This is the path's capstone. Copy the prompt into ChatGPT, Claude, Gemini, Copilot, or whichever AI tool you have access to — the website doesn't run AI itself. Answer its questions honestly; at the end you'll have a completed process doc in the six-line format. Save it next to your prompt library — this doc is the artifact the whole path was building toward.

I've finished a course on AI workflows, and I'm documenting my single most repeatable AI process. Interview me one question at a time — no more than eight questions total — covering: what the process is for, what triggers a run, the steps in order (including where the input comes from and what shape it's in), the exact prompt I use, what I review before trusting the output, and who owns keeping this current. If an answer is vague, push once for specifics. Then output my completed process doc, ready to save, in exactly this format:

PROCESS: (name that says the job)
Purpose:
Trigger:
Steps: (numbered)
Prompt: (my full prompt, complete)
Review:
Owner:
Last updated: (today's date)

Reflection: Save the doc where you'll find it. You now have a prompt library, a v1 system, and one documented process. Run the process next week — and when a run disappoints you, that's not failure, that's v2.

Quick check

4 quick questions — no pressure

There's no pass or fail here. Answer them all, and we'll show you the answers either way.